Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
f88c5d90 LE |
26 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
27 | ||
b411b363 PR |
28 | #include <linux/module.h> |
29 | #include <linux/drbd.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/fs.h> | |
32 | #include <linux/file.h> | |
33 | #include <linux/slab.h> | |
b411b363 PR |
34 | #include <linux/blkpg.h> |
35 | #include <linux/cpumask.h> | |
36 | #include "drbd_int.h" | |
a3603a6e | 37 | #include "drbd_protocol.h" |
265be2d0 | 38 | #include "drbd_req.h" |
a2972846 | 39 | #include "drbd_state_change.h" |
b411b363 | 40 | #include <asm/unaligned.h> |
b411b363 | 41 | #include <linux/drbd_limits.h> |
87f7be4c | 42 | #include <linux/kthread.h> |
b411b363 | 43 | |
3b98c0c2 LE |
44 | #include <net/genetlink.h> |
45 | ||
46 | /* .doit */ | |
47 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
48 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
05a10ec7 AG |
50 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 52 | |
789c1b62 AG |
53 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
54 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 55 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
56 | |
57 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
58 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 62 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
63 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
64 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
72 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
73 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 74 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
75 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
76 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
77 | /* .dumpit */ | |
78 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
a55bbd37 AG |
79 | int drbd_adm_dump_resources(struct sk_buff *skb, struct netlink_callback *cb); |
80 | int drbd_adm_dump_devices(struct sk_buff *skb, struct netlink_callback *cb); | |
81 | int drbd_adm_dump_devices_done(struct netlink_callback *cb); | |
82 | int drbd_adm_dump_connections(struct sk_buff *skb, struct netlink_callback *cb); | |
83 | int drbd_adm_dump_connections_done(struct netlink_callback *cb); | |
84 | int drbd_adm_dump_peer_devices(struct sk_buff *skb, struct netlink_callback *cb); | |
85 | int drbd_adm_dump_peer_devices_done(struct netlink_callback *cb); | |
a2972846 | 86 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb); |
3b98c0c2 LE |
87 | |
88 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 89 | #include "drbd_nla.h" |
3b98c0c2 LE |
90 | #include <linux/genl_magic_func.h> |
91 | ||
a2972846 AG |
92 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
93 | static atomic_t notify_genl_seq = ATOMIC_INIT(2); /* two. */ | |
94 | ||
95 | DEFINE_MUTEX(notification_mutex); | |
96 | ||
3b98c0c2 | 97 | /* used blkdev_get_by_path, to claim our meta data device(s) */ |
b411b363 PR |
98 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
99 | ||
3b98c0c2 LE |
100 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) |
101 | { | |
102 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
103 | if (genlmsg_reply(skb, info)) | |
f88c5d90 | 104 | pr_err("error sending genl reply\n"); |
b411b363 | 105 | } |
3b98c0c2 LE |
106 | |
107 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
108 | * reason it could fail was no space in skb, and there are 4k available. */ | |
f221f4bc | 109 | static int drbd_msg_put_info(struct sk_buff *skb, const char *info) |
3b98c0c2 | 110 | { |
3b98c0c2 LE |
111 | struct nlattr *nla; |
112 | int err = -EMSGSIZE; | |
113 | ||
114 | if (!info || !info[0]) | |
115 | return 0; | |
116 | ||
117 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
118 | if (!nla) | |
119 | return err; | |
120 | ||
121 | err = nla_put_string(skb, T_info_text, info); | |
122 | if (err) { | |
123 | nla_nest_cancel(skb, nla); | |
124 | return err; | |
125 | } else | |
126 | nla_nest_end(skb, nla); | |
127 | return 0; | |
b411b363 PR |
128 | } |
129 | ||
3b98c0c2 LE |
130 | /* This would be a good candidate for a "pre_doit" hook, |
131 | * and per-family private info->pointers. | |
132 | * But we need to stay compatible with older kernels. | |
133 | * If it returns successfully, adm_ctx members are valid. | |
9e276872 LE |
134 | * |
135 | * At this point, we still rely on the global genl_lock(). | |
136 | * If we want to avoid that, and allow "genl_family.parallel_ops", we may need | |
137 | * to add additional synchronization against object destruction/modification. | |
3b98c0c2 LE |
138 | */ |
139 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 140 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 141 | #define DRBD_ADM_NEED_CONNECTION 4 |
a910b123 LE |
142 | static int drbd_adm_prepare(struct drbd_config_context *adm_ctx, |
143 | struct sk_buff *skb, struct genl_info *info, unsigned flags) | |
3b98c0c2 LE |
144 | { |
145 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
146 | const u8 cmd = info->genlhdr->cmd; | |
147 | int err; | |
148 | ||
a910b123 | 149 | memset(adm_ctx, 0, sizeof(*adm_ctx)); |
3b98c0c2 LE |
150 | |
151 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 152 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
153 | return -EPERM; |
154 | ||
a910b123 LE |
155 | adm_ctx->reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); |
156 | if (!adm_ctx->reply_skb) { | |
1e2a2551 | 157 | err = -ENOMEM; |
3b98c0c2 | 158 | goto fail; |
1e2a2551 | 159 | } |
3b98c0c2 | 160 | |
a910b123 | 161 | adm_ctx->reply_dh = genlmsg_put_reply(adm_ctx->reply_skb, |
3b98c0c2 LE |
162 | info, &drbd_genl_family, 0, cmd); |
163 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
164 | * but anyways */ | |
a910b123 | 165 | if (!adm_ctx->reply_dh) { |
1e2a2551 | 166 | err = -ENOMEM; |
3b98c0c2 | 167 | goto fail; |
1e2a2551 | 168 | } |
3b98c0c2 | 169 | |
a910b123 LE |
170 | adm_ctx->reply_dh->minor = d_in->minor; |
171 | adm_ctx->reply_dh->ret_code = NO_ERROR; | |
3b98c0c2 | 172 | |
a910b123 | 173 | adm_ctx->volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
174 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
175 | struct nlattr *nla; | |
176 | /* parse and validate only */ | |
f399002e | 177 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
178 | if (err) |
179 | goto fail; | |
180 | ||
181 | /* It was present, and valid, | |
182 | * copy it over to the reply skb. */ | |
a910b123 | 183 | err = nla_put_nohdr(adm_ctx->reply_skb, |
3b98c0c2 LE |
184 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, |
185 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
186 | if (err) | |
187 | goto fail; | |
188 | ||
9e276872 | 189 | /* and assign stuff to the adm_ctx */ |
3b98c0c2 | 190 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; |
089c075d | 191 | if (nla) |
a910b123 | 192 | adm_ctx->volume = nla_get_u32(nla); |
7c3063cc | 193 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 194 | if (nla) |
a910b123 LE |
195 | adm_ctx->resource_name = nla_data(nla); |
196 | adm_ctx->my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; | |
197 | adm_ctx->peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
198 | if ((adm_ctx->my_addr && | |
199 | nla_len(adm_ctx->my_addr) > sizeof(adm_ctx->connection->my_addr)) || | |
200 | (adm_ctx->peer_addr && | |
201 | nla_len(adm_ctx->peer_addr) > sizeof(adm_ctx->connection->peer_addr))) { | |
089c075d AG |
202 | err = -EINVAL; |
203 | goto fail; | |
204 | } | |
205 | } | |
3b98c0c2 | 206 | |
a910b123 LE |
207 | adm_ctx->minor = d_in->minor; |
208 | adm_ctx->device = minor_to_device(d_in->minor); | |
9e276872 LE |
209 | |
210 | /* We are protected by the global genl_lock(). | |
211 | * But we may explicitly drop it/retake it in drbd_adm_set_role(), | |
212 | * so make sure this object stays around. */ | |
213 | if (adm_ctx->device) | |
214 | kref_get(&adm_ctx->device->kref); | |
215 | ||
a910b123 LE |
216 | if (adm_ctx->resource_name) { |
217 | adm_ctx->resource = drbd_find_resource(adm_ctx->resource_name); | |
4bc76048 | 218 | } |
3b98c0c2 | 219 | |
a910b123 LE |
220 | if (!adm_ctx->device && (flags & DRBD_ADM_NEED_MINOR)) { |
221 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown minor"); | |
3b98c0c2 LE |
222 | return ERR_MINOR_INVALID; |
223 | } | |
a910b123 LE |
224 | if (!adm_ctx->resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
225 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown resource"); | |
226 | if (adm_ctx->resource_name) | |
a10f6b8a | 227 | return ERR_RES_NOT_KNOWN; |
3b98c0c2 LE |
228 | return ERR_INVALID_REQUEST; |
229 | } | |
230 | ||
089c075d | 231 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
a910b123 LE |
232 | if (adm_ctx->resource) { |
233 | drbd_msg_put_info(adm_ctx->reply_skb, "no resource name expected"); | |
089c075d AG |
234 | return ERR_INVALID_REQUEST; |
235 | } | |
a910b123 LE |
236 | if (adm_ctx->device) { |
237 | drbd_msg_put_info(adm_ctx->reply_skb, "no minor number expected"); | |
089c075d AG |
238 | return ERR_INVALID_REQUEST; |
239 | } | |
a910b123 LE |
240 | if (adm_ctx->my_addr && adm_ctx->peer_addr) |
241 | adm_ctx->connection = conn_get_by_addrs(nla_data(adm_ctx->my_addr), | |
242 | nla_len(adm_ctx->my_addr), | |
243 | nla_data(adm_ctx->peer_addr), | |
244 | nla_len(adm_ctx->peer_addr)); | |
245 | if (!adm_ctx->connection) { | |
246 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown connection"); | |
089c075d AG |
247 | return ERR_INVALID_REQUEST; |
248 | } | |
249 | } | |
250 | ||
3b98c0c2 | 251 | /* some more paranoia, if the request was over-determined */ |
a910b123 LE |
252 | if (adm_ctx->device && adm_ctx->resource && |
253 | adm_ctx->device->resource != adm_ctx->resource) { | |
4bc76048 | 254 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", |
a910b123 LE |
255 | adm_ctx->minor, adm_ctx->resource->name, |
256 | adm_ctx->device->resource->name); | |
257 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists in different resource"); | |
527f4b24 LE |
258 | return ERR_INVALID_REQUEST; |
259 | } | |
a910b123 LE |
260 | if (adm_ctx->device && |
261 | adm_ctx->volume != VOLUME_UNSPECIFIED && | |
262 | adm_ctx->volume != adm_ctx->device->vnr) { | |
3b98c0c2 | 263 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
a910b123 LE |
264 | adm_ctx->minor, adm_ctx->volume, |
265 | adm_ctx->device->vnr, | |
266 | adm_ctx->device->resource->name); | |
267 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists as different volume"); | |
3b98c0c2 LE |
268 | return ERR_INVALID_REQUEST; |
269 | } | |
0ace9dfa | 270 | |
9e276872 LE |
271 | /* still, provide adm_ctx->resource always, if possible. */ |
272 | if (!adm_ctx->resource) { | |
273 | adm_ctx->resource = adm_ctx->device ? adm_ctx->device->resource | |
274 | : adm_ctx->connection ? adm_ctx->connection->resource : NULL; | |
275 | if (adm_ctx->resource) | |
276 | kref_get(&adm_ctx->resource->kref); | |
277 | } | |
278 | ||
3b98c0c2 LE |
279 | return NO_ERROR; |
280 | ||
281 | fail: | |
a910b123 LE |
282 | nlmsg_free(adm_ctx->reply_skb); |
283 | adm_ctx->reply_skb = NULL; | |
1e2a2551 | 284 | return err; |
3b98c0c2 LE |
285 | } |
286 | ||
a910b123 LE |
287 | static int drbd_adm_finish(struct drbd_config_context *adm_ctx, |
288 | struct genl_info *info, int retcode) | |
3b98c0c2 | 289 | { |
9e276872 LE |
290 | if (adm_ctx->device) { |
291 | kref_put(&adm_ctx->device->kref, drbd_destroy_device); | |
292 | adm_ctx->device = NULL; | |
293 | } | |
a910b123 LE |
294 | if (adm_ctx->connection) { |
295 | kref_put(&adm_ctx->connection->kref, &drbd_destroy_connection); | |
296 | adm_ctx->connection = NULL; | |
0ace9dfa | 297 | } |
a910b123 LE |
298 | if (adm_ctx->resource) { |
299 | kref_put(&adm_ctx->resource->kref, drbd_destroy_resource); | |
300 | adm_ctx->resource = NULL; | |
4bc76048 | 301 | } |
0ace9dfa | 302 | |
a910b123 | 303 | if (!adm_ctx->reply_skb) |
3b98c0c2 LE |
304 | return -ENOMEM; |
305 | ||
a910b123 LE |
306 | adm_ctx->reply_dh->ret_code = retcode; |
307 | drbd_adm_send_reply(adm_ctx->reply_skb, info); | |
3b98c0c2 LE |
308 | return 0; |
309 | } | |
b411b363 | 310 | |
bde89a9e | 311 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 312 | { |
6b75dced | 313 | char *afs; |
b411b363 | 314 | |
089c075d | 315 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 316 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
317 | return; |
318 | ||
bde89a9e | 319 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
320 | case AF_INET6: |
321 | afs = "ipv6"; | |
322 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 323 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 324 | break; |
089c075d AG |
325 | case AF_INET: |
326 | afs = "ipv4"; | |
327 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 328 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 329 | break; |
089c075d AG |
330 | default: |
331 | afs = "ssocks"; | |
332 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 333 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 334 | } |
089c075d | 335 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 336 | } |
b411b363 | 337 | |
b30ab791 | 338 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
339 | { |
340 | char *envp[] = { "HOME=/", | |
341 | "TERM=linux", | |
342 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
343 | (char[20]) { }, /* address family */ |
344 | (char[60]) { }, /* address */ | |
b411b363 | 345 | NULL }; |
0982368b | 346 | char mb[14]; |
8ab761e1 | 347 | char *argv[] = {drbd_usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 348 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 349 | struct sib_info sib; |
b411b363 PR |
350 | int ret; |
351 | ||
bde89a9e AG |
352 | if (current == connection->worker.task) |
353 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 354 | |
0982368b | 355 | snprintf(mb, 14, "minor-%d", device_to_minor(device)); |
bde89a9e | 356 | setup_khelper_env(connection, envp); |
b411b363 | 357 | |
1090c056 LE |
358 | /* The helper may take some time. |
359 | * write out any unsynced meta data changes now */ | |
b30ab791 | 360 | drbd_md_sync(device); |
1090c056 | 361 | |
8ab761e1 | 362 | drbd_info(device, "helper command: %s %s %s\n", drbd_usermode_helper, cmd, mb); |
3b98c0c2 LE |
363 | sib.sib_reason = SIB_HELPER_PRE; |
364 | sib.helper_name = cmd; | |
b30ab791 | 365 | drbd_bcast_event(device, &sib); |
a2972846 | 366 | notify_helper(NOTIFY_CALL, device, connection, cmd, 0); |
8ab761e1 | 367 | ret = call_usermodehelper(drbd_usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 368 | if (ret) |
d0180171 | 369 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
8ab761e1 | 370 | drbd_usermode_helper, cmd, mb, |
b411b363 PR |
371 | (ret >> 8) & 0xff, ret); |
372 | else | |
d0180171 | 373 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
8ab761e1 | 374 | drbd_usermode_helper, cmd, mb, |
b411b363 | 375 | (ret >> 8) & 0xff, ret); |
3b98c0c2 LE |
376 | sib.sib_reason = SIB_HELPER_POST; |
377 | sib.helper_exit_code = ret; | |
b30ab791 | 378 | drbd_bcast_event(device, &sib); |
a2972846 | 379 | notify_helper(NOTIFY_RESPONSE, device, connection, cmd, ret); |
b411b363 | 380 | |
bde89a9e AG |
381 | if (current == connection->worker.task) |
382 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
383 | |
384 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
385 | ret = 0; | |
386 | ||
387 | return ret; | |
388 | } | |
389 | ||
7e5fec31 | 390 | enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
391 | { |
392 | char *envp[] = { "HOME=/", | |
393 | "TERM=linux", | |
394 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
395 | (char[20]) { }, /* address family */ | |
396 | (char[60]) { }, /* address */ | |
397 | NULL }; | |
77c556f6 | 398 | char *resource_name = connection->resource->name; |
8ab761e1 | 399 | char *argv[] = {drbd_usermode_helper, cmd, resource_name, NULL }; |
6b75dced PR |
400 | int ret; |
401 | ||
bde89a9e AG |
402 | setup_khelper_env(connection, envp); |
403 | conn_md_sync(connection); | |
6b75dced | 404 | |
8ab761e1 | 405 | drbd_info(connection, "helper command: %s %s %s\n", drbd_usermode_helper, cmd, resource_name); |
6b75dced | 406 | /* TODO: conn_bcast_event() ?? */ |
a2972846 | 407 | notify_helper(NOTIFY_CALL, NULL, connection, cmd, 0); |
6b75dced | 408 | |
8ab761e1 | 409 | ret = call_usermodehelper(drbd_usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 410 | if (ret) |
1ec861eb | 411 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
8ab761e1 | 412 | drbd_usermode_helper, cmd, resource_name, |
6b75dced PR |
413 | (ret >> 8) & 0xff, ret); |
414 | else | |
1ec861eb | 415 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
8ab761e1 | 416 | drbd_usermode_helper, cmd, resource_name, |
6b75dced PR |
417 | (ret >> 8) & 0xff, ret); |
418 | /* TODO: conn_bcast_event() ?? */ | |
a2972846 | 419 | notify_helper(NOTIFY_RESPONSE, NULL, connection, cmd, ret); |
c2ba686f | 420 | |
b411b363 PR |
421 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
422 | ret = 0; | |
423 | ||
424 | return ret; | |
425 | } | |
426 | ||
bde89a9e | 427 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 428 | { |
cb703454 | 429 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 430 | struct drbd_peer_device *peer_device; |
cb703454 PR |
431 | int vnr; |
432 | ||
695d08fa | 433 | rcu_read_lock(); |
c06ece6b AG |
434 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
435 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 436 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
437 | struct disk_conf *disk_conf = |
438 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
439 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 440 | put_ldev(device); |
cb703454 PR |
441 | } |
442 | } | |
695d08fa | 443 | rcu_read_unlock(); |
cb703454 PR |
444 | |
445 | return fp; | |
446 | } | |
447 | ||
7bd000cb LE |
448 | static bool resource_is_supended(struct drbd_resource *resource) |
449 | { | |
450 | return resource->susp || resource->susp_fen || resource->susp_nod; | |
451 | } | |
452 | ||
bde89a9e | 453 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 454 | { |
7bd000cb | 455 | struct drbd_resource * const resource = connection->resource; |
28e448bb | 456 | unsigned int connect_cnt; |
cb703454 PR |
457 | union drbd_state mask = { }; |
458 | union drbd_state val = { }; | |
459 | enum drbd_fencing_p fp; | |
b411b363 PR |
460 | char *ex_to_string; |
461 | int r; | |
b411b363 | 462 | |
7bd000cb | 463 | spin_lock_irq(&resource->req_lock); |
bde89a9e | 464 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 465 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
7bd000cb | 466 | spin_unlock_irq(&resource->req_lock); |
cb703454 PR |
467 | return false; |
468 | } | |
b411b363 | 469 | |
bde89a9e | 470 | connect_cnt = connection->connect_cnt; |
7bd000cb | 471 | spin_unlock_irq(&resource->req_lock); |
28e448bb | 472 | |
bde89a9e | 473 | fp = highest_fencing_policy(connection); |
cb703454 PR |
474 | switch (fp) { |
475 | case FP_NOT_AVAIL: | |
1ec861eb | 476 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
7bd000cb LE |
477 | spin_lock_irq(&resource->req_lock); |
478 | if (connection->cstate < C_WF_REPORT_PARAMS) { | |
479 | _conn_request_state(connection, | |
480 | (union drbd_state) { { .susp_fen = 1 } }, | |
481 | (union drbd_state) { { .susp_fen = 0 } }, | |
482 | CS_VERBOSE | CS_HARD | CS_DC_SUSP); | |
483 | /* We are no longer suspended due to the fencing policy. | |
484 | * We may still be suspended due to the on-no-data-accessible policy. | |
485 | * If that was OND_IO_ERROR, fail pending requests. */ | |
486 | if (!resource_is_supended(resource)) | |
487 | _tl_restart(connection, CONNECTION_LOST_WHILE_PENDING); | |
488 | } | |
489 | /* Else: in case we raced with a connection handshake, | |
490 | * let the handshake figure out if we maybe can RESEND, | |
491 | * and do not resume/fail pending requests here. | |
492 | * Worst case is we stay suspended for now, which may be | |
493 | * resolved by either re-establishing the replication link, or | |
494 | * the next link failure, or eventually the administrator. */ | |
495 | spin_unlock_irq(&resource->req_lock); | |
496 | return false; | |
497 | ||
cb703454 PR |
498 | case FP_DONT_CARE: |
499 | return true; | |
500 | default: ; | |
b411b363 PR |
501 | } |
502 | ||
bde89a9e | 503 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
504 | |
505 | switch ((r>>8) & 0xff) { | |
7e5fec31 | 506 | case P_INCONSISTENT: /* peer is inconsistent */ |
b411b363 | 507 | ex_to_string = "peer is inconsistent or worse"; |
cb703454 PR |
508 | mask.pdsk = D_MASK; |
509 | val.pdsk = D_INCONSISTENT; | |
b411b363 | 510 | break; |
7e5fec31 | 511 | case P_OUTDATED: /* peer got outdated, or was already outdated */ |
b411b363 | 512 | ex_to_string = "peer was fenced"; |
cb703454 PR |
513 | mask.pdsk = D_MASK; |
514 | val.pdsk = D_OUTDATED; | |
b411b363 | 515 | break; |
7e5fec31 | 516 | case P_DOWN: /* peer was down */ |
bde89a9e | 517 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
518 | /* we will(have) create(d) a new UUID anyways... */ |
519 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
520 | mask.pdsk = D_MASK; |
521 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
522 | } else { |
523 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
524 | } |
525 | break; | |
7e5fec31 | 526 | case P_PRIMARY: /* Peer is primary, voluntarily outdate myself. |
b411b363 PR |
527 | * This is useful when an unconnected R_SECONDARY is asked to |
528 | * become R_PRIMARY, but finds the other peer being active. */ | |
529 | ex_to_string = "peer is active"; | |
1ec861eb | 530 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
531 | mask.disk = D_MASK; |
532 | val.disk = D_OUTDATED; | |
b411b363 | 533 | break; |
7e5fec31 FF |
534 | case P_FENCING: |
535 | /* THINK: do we need to handle this | |
536 | * like case 4, or more like case 5? */ | |
b411b363 | 537 | if (fp != FP_STONITH) |
1ec861eb | 538 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 539 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
540 | mask.pdsk = D_MASK; |
541 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
542 | break; |
543 | default: | |
544 | /* The script is broken ... */ | |
1ec861eb | 545 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 546 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
547 | } |
548 | ||
1ec861eb | 549 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 550 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 551 | |
cb703454 | 552 | /* Not using |
bde89a9e | 553 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
554 | here, because we might were able to re-establish the connection in the |
555 | meantime. */ | |
7bd000cb | 556 | spin_lock_irq(&resource->req_lock); |
bde89a9e AG |
557 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
558 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
559 | /* In case the connection was established and droped |
560 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 561 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 562 | else |
bde89a9e | 563 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 564 | } |
7bd000cb | 565 | spin_unlock_irq(&resource->req_lock); |
cb703454 | 566 | |
bde89a9e | 567 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
568 | } |
569 | ||
87f7be4c PR |
570 | static int _try_outdate_peer_async(void *data) |
571 | { | |
bde89a9e | 572 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 573 | |
bde89a9e | 574 | conn_try_outdate_peer(connection); |
87f7be4c | 575 | |
05a10ec7 | 576 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
577 | return 0; |
578 | } | |
579 | ||
bde89a9e | 580 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
581 | { |
582 | struct task_struct *opa; | |
583 | ||
bde89a9e | 584 | kref_get(&connection->kref); |
bbc1c5e8 LE |
585 | /* We may just have force_sig()'ed this thread |
586 | * to get it out of some blocking network function. | |
587 | * Clear signals; otherwise kthread_run(), which internally uses | |
588 | * wait_on_completion_killable(), will mistake our pending signal | |
589 | * for a new fatal signal and fail. */ | |
590 | flush_signals(current); | |
bde89a9e | 591 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); |
9dc9fbb3 | 592 | if (IS_ERR(opa)) { |
1ec861eb | 593 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 594 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 595 | } |
87f7be4c | 596 | } |
b411b363 | 597 | |
bf885f8a | 598 | enum drbd_state_rv |
44a4d551 | 599 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 600 | { |
44a4d551 LE |
601 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
602 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 603 | const int max_tries = 4; |
bf885f8a | 604 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 605 | struct net_conf *nc; |
b411b363 PR |
606 | int try = 0; |
607 | int forced = 0; | |
608 | union drbd_state mask, val; | |
b411b363 | 609 | |
b6f85ef9 AG |
610 | if (new_role == R_PRIMARY) { |
611 | struct drbd_connection *connection; | |
612 | ||
613 | /* Detect dead peers as soon as possible. */ | |
614 | ||
615 | rcu_read_lock(); | |
616 | for_each_connection(connection, device->resource) | |
617 | request_ping(connection); | |
618 | rcu_read_unlock(); | |
619 | } | |
b411b363 | 620 | |
b30ab791 | 621 | mutex_lock(device->state_mutex); |
b411b363 PR |
622 | |
623 | mask.i = 0; mask.role = R_MASK; | |
624 | val.i = 0; val.role = new_role; | |
625 | ||
626 | while (try++ < max_tries) { | |
a8821531 | 627 | rv = _drbd_request_state_holding_state_mutex(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
628 | |
629 | /* in case we first succeeded to outdate, | |
630 | * but now suddenly could establish a connection */ | |
bf885f8a | 631 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
632 | val.pdsk = 0; |
633 | mask.pdsk = 0; | |
634 | continue; | |
635 | } | |
636 | ||
bf885f8a | 637 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
638 | (device->state.disk < D_UP_TO_DATE && |
639 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
640 | mask.disk = D_MASK; |
641 | val.disk = D_UP_TO_DATE; | |
642 | forced = 1; | |
643 | continue; | |
644 | } | |
645 | ||
bf885f8a | 646 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 647 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 648 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 649 | |
44a4d551 | 650 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
651 | val.disk = D_UP_TO_DATE; |
652 | mask.disk = D_MASK; | |
653 | } | |
b411b363 PR |
654 | continue; |
655 | } | |
656 | ||
bf885f8a | 657 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 658 | goto out; |
bf885f8a | 659 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 660 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 661 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
662 | mask.pdsk = D_MASK; |
663 | val.pdsk = D_OUTDATED; | |
b411b363 | 664 | |
cb703454 | 665 | } |
b411b363 PR |
666 | continue; |
667 | } | |
bf885f8a | 668 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
669 | /* Maybe the peer is detected as dead very soon... |
670 | retry at most once more in this case. */ | |
44ed167d PR |
671 | int timeo; |
672 | rcu_read_lock(); | |
44a4d551 | 673 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
674 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
675 | rcu_read_unlock(); | |
676 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
677 | if (try < max_tries) |
678 | try = max_tries - 1; | |
679 | continue; | |
680 | } | |
bf885f8a | 681 | if (rv < SS_SUCCESS) { |
b30ab791 | 682 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 683 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 684 | if (rv < SS_SUCCESS) |
3b98c0c2 | 685 | goto out; |
b411b363 PR |
686 | } |
687 | break; | |
688 | } | |
689 | ||
bf885f8a | 690 | if (rv < SS_SUCCESS) |
3b98c0c2 | 691 | goto out; |
b411b363 PR |
692 | |
693 | if (forced) | |
d0180171 | 694 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
695 | |
696 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 697 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 698 | |
b6dd1a89 LE |
699 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
700 | ||
b411b363 | 701 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
702 | if (get_ldev(device)) { |
703 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
704 | put_ldev(device); | |
b411b363 PR |
705 | } |
706 | } else { | |
66ce6dbc | 707 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 708 | nc = connection->net_conf; |
44ed167d | 709 | if (nc) |
6139f60d | 710 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 711 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 712 | |
b30ab791 AG |
713 | if (get_ldev(device)) { |
714 | if (((device->state.conn < C_CONNECTED || | |
715 | device->state.pdsk <= D_FAILED) | |
716 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
717 | drbd_uuid_new_current(device); | |
b411b363 | 718 | |
b30ab791 AG |
719 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
720 | put_ldev(device); | |
b411b363 PR |
721 | } |
722 | } | |
723 | ||
19f843aa LE |
724 | /* writeout of activity log covered areas of the bitmap |
725 | * to stable storage done in after state change already */ | |
b411b363 | 726 | |
b30ab791 | 727 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
728 | /* if this was forced, we should consider sync */ |
729 | if (forced) | |
44a4d551 LE |
730 | drbd_send_uuids(peer_device); |
731 | drbd_send_current_state(peer_device); | |
b411b363 PR |
732 | } |
733 | ||
b30ab791 | 734 | drbd_md_sync(device); |
720979fb | 735 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 736 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 737 | out: |
b30ab791 | 738 | mutex_unlock(device->state_mutex); |
bf885f8a | 739 | return rv; |
b411b363 PR |
740 | } |
741 | ||
3b98c0c2 | 742 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 743 | { |
3b98c0c2 LE |
744 | return err == -ENOMSG ? "required attribute missing" : |
745 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 746 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 747 | "invalid attribute value"; |
ef50a3e3 | 748 | } |
b411b363 | 749 | |
3b98c0c2 | 750 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 751 | { |
a910b123 | 752 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
753 | struct set_role_parms parms; |
754 | int err; | |
755 | enum drbd_ret_code retcode; | |
b411b363 | 756 | |
a910b123 | 757 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
758 | if (!adm_ctx.reply_skb) |
759 | return retcode; | |
760 | if (retcode != NO_ERROR) | |
761 | goto out; | |
b411b363 | 762 | |
3b98c0c2 LE |
763 | memset(&parms, 0, sizeof(parms)); |
764 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 765 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
766 | if (err) { |
767 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 768 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
769 | goto out; |
770 | } | |
771 | } | |
9e276872 LE |
772 | genl_unlock(); |
773 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 774 | |
3b98c0c2 | 775 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 776 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 777 | else |
b30ab791 | 778 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
779 | |
780 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
781 | genl_lock(); | |
3b98c0c2 | 782 | out: |
a910b123 | 783 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
784 | return 0; |
785 | } | |
786 | ||
ae8bf312 LE |
787 | /* Initializes the md.*_offset members, so we are able to find |
788 | * the on disk meta data. | |
789 | * | |
790 | * We currently have two possible layouts: | |
791 | * external: | |
792 | * |----------- md_size_sect ------------------| | |
793 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
794 | * | al_offset == 8 | | |
795 | * | bm_offset = al_offset + X | | |
796 | * ==> bitmap sectors = md_size_sect - bm_offset | |
797 | * | |
798 | * internal: | |
799 | * |----------- md_size_sect ------------------| | |
800 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
801 | * | al_offset < 0 | | |
802 | * | bm_offset = al_offset - Y | | |
803 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
804 | * | |
805 | * Activity log size used to be fixed 32kB, | |
806 | * but is about to become configurable. | |
807 | */ | |
b30ab791 | 808 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
809 | struct drbd_backing_dev *bdev) |
810 | { | |
811 | sector_t md_size_sect = 0; | |
c04ccaa6 | 812 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 813 | |
3a4d4eb3 LE |
814 | bdev->md.md_offset = drbd_md_ss(bdev); |
815 | ||
68e41a43 | 816 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
817 | default: |
818 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 819 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
820 | bdev->md.al_offset = MD_4kB_SECT; |
821 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
822 | break; |
823 | case DRBD_MD_INDEX_FLEX_EXT: | |
824 | /* just occupy the full device; unit: sectors */ | |
825 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
826 | bdev->md.al_offset = MD_4kB_SECT; |
827 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
828 | break; |
829 | case DRBD_MD_INDEX_INTERNAL: | |
830 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 831 | /* al size is still fixed */ |
ae8bf312 | 832 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
833 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
834 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
835 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
836 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
837 | md_size_sect = ALIGN(md_size_sect, 8); | |
838 | ||
839 | /* plus the "drbd meta data super block", | |
840 | * and the activity log; */ | |
ae8bf312 | 841 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
842 | |
843 | bdev->md.md_size_sect = md_size_sect; | |
844 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 845 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
846 | break; |
847 | } | |
848 | } | |
849 | ||
4b0715f0 | 850 | /* input size is expected to be in KB */ |
b411b363 PR |
851 | char *ppsize(char *buf, unsigned long long size) |
852 | { | |
4b0715f0 LE |
853 | /* Needs 9 bytes at max including trailing NUL: |
854 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
855 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
856 | int base = 0; | |
4b0715f0 | 857 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
858 | /* shift + round */ |
859 | size = (size >> 10) + !!(size & (1<<9)); | |
860 | base++; | |
861 | } | |
4b0715f0 | 862 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
863 | |
864 | return buf; | |
865 | } | |
866 | ||
867 | /* there is still a theoretical deadlock when called from receiver | |
868 | * on an D_INCONSISTENT R_PRIMARY: | |
869 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
870 | * packet from remote to dec_ap_bio again. | |
871 | * receiver receive_sizes(), comes here, | |
872 | * waits for ap_bio_cnt == 0. -> deadlock. | |
873 | * but this cannot happen, actually, because: | |
874 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
875 | * (not connected, or bad/no disk on peer): | |
876 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
877 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
878 | * peer may not initiate a resize. | |
879 | */ | |
3b98c0c2 LE |
880 | /* Note these are not to be confused with |
881 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
882 | * which are (sub) state changes triggered by admin (drbdsetup), | |
883 | * and can be long lived. | |
b30ab791 | 884 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 885 | * and should be short-lived. */ |
7dbb4386 PR |
886 | /* It needs to be a counter, since multiple threads might |
887 | independently suspend and resume IO. */ | |
b30ab791 | 888 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 889 | { |
7dbb4386 | 890 | atomic_inc(&device->suspend_cnt); |
b30ab791 | 891 | if (drbd_suspended(device)) |
265be2d0 | 892 | return; |
b30ab791 | 893 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
894 | } |
895 | ||
b30ab791 | 896 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 897 | { |
7dbb4386 PR |
898 | if (atomic_dec_and_test(&device->suspend_cnt)) |
899 | wake_up(&device->misc_wait); | |
b411b363 PR |
900 | } |
901 | ||
902 | /** | |
903 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 904 | * @device: DRBD device. |
b411b363 PR |
905 | * |
906 | * Returns 0 on success, negative return values indicate errors. | |
907 | * You should call drbd_md_sync() after calling this function. | |
908 | */ | |
d752b269 | 909 | enum determine_dev_size |
b30ab791 | 910 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 | 911 | { |
8011e249 LE |
912 | struct md_offsets_and_sizes { |
913 | u64 last_agreed_sect; | |
914 | u64 md_offset; | |
915 | s32 al_offset; | |
916 | s32 bm_offset; | |
917 | u32 md_size_sect; | |
918 | ||
919 | u32 al_stripes; | |
920 | u32 al_stripe_size_4k; | |
921 | } prev; | |
922 | sector_t u_size, size; | |
b30ab791 | 923 | struct drbd_md *md = &device->ldev->md; |
b411b363 | 924 | char ppb[10]; |
d752b269 | 925 | void *buffer; |
b411b363 PR |
926 | |
927 | int md_moved, la_size_changed; | |
e96c9633 | 928 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 | 929 | |
5f7c0124 LE |
930 | /* We may change the on-disk offsets of our meta data below. Lock out |
931 | * anything that may cause meta data IO, to avoid acting on incomplete | |
932 | * layout changes or scribbling over meta data that is in the process | |
933 | * of being moved. | |
b411b363 | 934 | * |
5f7c0124 LE |
935 | * Move is not exactly correct, btw, currently we have all our meta |
936 | * data in core memory, to "move" it we just write it all out, there | |
937 | * are no reads. */ | |
b30ab791 | 938 | drbd_suspend_io(device); |
e37d2438 | 939 | buffer = drbd_md_get_buffer(device, __func__); /* Lock meta-data IO */ |
d752b269 | 940 | if (!buffer) { |
b30ab791 | 941 | drbd_resume_io(device); |
d752b269 PR |
942 | return DS_ERROR; |
943 | } | |
b411b363 | 944 | |
8011e249 LE |
945 | /* remember current offset and sizes */ |
946 | prev.last_agreed_sect = md->la_size_sect; | |
947 | prev.md_offset = md->md_offset; | |
948 | prev.al_offset = md->al_offset; | |
949 | prev.bm_offset = md->bm_offset; | |
950 | prev.md_size_sect = md->md_size_sect; | |
951 | prev.al_stripes = md->al_stripes; | |
952 | prev.al_stripe_size_4k = md->al_stripe_size_4k; | |
b411b363 | 953 | |
d752b269 PR |
954 | if (rs) { |
955 | /* rs is non NULL if we should change the AL layout only */ | |
d752b269 PR |
956 | md->al_stripes = rs->al_stripes; |
957 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
958 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
959 | } | |
960 | ||
b30ab791 | 961 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 962 | |
daeda1cc | 963 | rcu_read_lock(); |
b30ab791 | 964 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 965 | rcu_read_unlock(); |
b30ab791 | 966 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 967 | |
8011e249 | 968 | if (size < prev.last_agreed_sect) { |
d752b269 PR |
969 | if (rs && u_size == 0) { |
970 | /* Remove "rs &&" later. This check should always be active, but | |
971 | right now the receiver expects the permissive behavior */ | |
d0180171 | 972 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
973 | "Use --size=%llus for explicit shrink.\n", |
974 | (unsigned long long)size); | |
975 | rv = DS_ERROR_SHRINK; | |
976 | } | |
977 | if (u_size > size) | |
978 | rv = DS_ERROR_SPACE_MD; | |
979 | if (rv != DS_UNCHANGED) | |
980 | goto err_out; | |
981 | } | |
982 | ||
b30ab791 AG |
983 | if (drbd_get_capacity(device->this_bdev) != size || |
984 | drbd_bm_capacity(device) != size) { | |
b411b363 | 985 | int err; |
b30ab791 | 986 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
987 | if (unlikely(err)) { |
988 | /* currently there is only one error: ENOMEM! */ | |
8011e249 | 989 | size = drbd_bm_capacity(device); |
b411b363 | 990 | if (size == 0) { |
d0180171 | 991 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
992 | "Could not allocate bitmap!\n"); |
993 | } else { | |
d0180171 | 994 | drbd_err(device, "BM resizing failed. " |
8011e249 | 995 | "Leaving size unchanged\n"); |
b411b363 | 996 | } |
e96c9633 | 997 | rv = DS_ERROR; |
b411b363 PR |
998 | } |
999 | /* racy, see comments above. */ | |
b30ab791 | 1000 | drbd_set_my_capacity(device, size); |
8011e249 | 1001 | md->la_size_sect = size; |
d0180171 | 1002 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
1003 | (unsigned long long)size>>1); |
1004 | } | |
d752b269 PR |
1005 | if (rv <= DS_ERROR) |
1006 | goto err_out; | |
b411b363 | 1007 | |
8011e249 | 1008 | la_size_changed = (prev.last_agreed_sect != md->la_size_sect); |
b411b363 | 1009 | |
8011e249 LE |
1010 | md_moved = prev.md_offset != md->md_offset |
1011 | || prev.md_size_sect != md->md_size_sect; | |
b411b363 | 1012 | |
d752b269 PR |
1013 | if (la_size_changed || md_moved || rs) { |
1014 | u32 prev_flags; | |
24dccabb | 1015 | |
fcb09674 LE |
1016 | /* We do some synchronous IO below, which may take some time. |
1017 | * Clear the timer, to avoid scary "timer expired!" messages, | |
1018 | * "Superblock" is written out at least twice below, anyways. */ | |
1019 | del_timer(&device->md_sync_timer); | |
d752b269 | 1020 | |
5f7c0124 LE |
1021 | /* We won't change the "al-extents" setting, we just may need |
1022 | * to move the on-disk location of the activity log ringbuffer. | |
1023 | * Lock for transaction is good enough, it may well be "dirty" | |
1024 | * or even "starving". */ | |
1025 | wait_event(device->al_wait, lc_try_lock_for_transaction(device->act_log)); | |
1026 | ||
1027 | /* mark current on-disk bitmap and activity log as unreliable */ | |
d752b269 | 1028 | prev_flags = md->flags; |
5f7c0124 | 1029 | md->flags |= MDF_FULL_SYNC | MDF_AL_DISABLED; |
b30ab791 | 1030 | drbd_md_write(device, buffer); |
d752b269 | 1031 | |
5f7c0124 LE |
1032 | drbd_al_initialize(device, buffer); |
1033 | ||
d0180171 | 1034 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
1035 | la_size_changed && md_moved ? "size changed and md moved" : |
1036 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 1037 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 1038 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 1039 | "size changed", BM_LOCKED_MASK); |
d752b269 | 1040 | |
5f7c0124 LE |
1041 | /* on-disk bitmap and activity log is authoritative again |
1042 | * (unless there was an IO error meanwhile...) */ | |
d752b269 | 1043 | md->flags = prev_flags; |
b30ab791 | 1044 | drbd_md_write(device, buffer); |
d752b269 PR |
1045 | |
1046 | if (rs) | |
d0180171 AG |
1047 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
1048 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
1049 | } |
1050 | ||
8011e249 LE |
1051 | if (size > prev.last_agreed_sect) |
1052 | rv = prev.last_agreed_sect ? DS_GREW : DS_GREW_FROM_ZERO; | |
1053 | if (size < prev.last_agreed_sect) | |
e96c9633 | 1054 | rv = DS_SHRUNK; |
d752b269 PR |
1055 | |
1056 | if (0) { | |
1057 | err_out: | |
8011e249 LE |
1058 | /* restore previous offset and sizes */ |
1059 | md->la_size_sect = prev.last_agreed_sect; | |
1060 | md->md_offset = prev.md_offset; | |
1061 | md->al_offset = prev.al_offset; | |
1062 | md->bm_offset = prev.bm_offset; | |
1063 | md->md_size_sect = prev.md_size_sect; | |
1064 | md->al_stripes = prev.al_stripes; | |
1065 | md->al_stripe_size_4k = prev.al_stripe_size_4k; | |
1066 | md->al_size_4k = (u64)prev.al_stripes * prev.al_stripe_size_4k; | |
d752b269 | 1067 | } |
b30ab791 AG |
1068 | lc_unlock(device->act_log); |
1069 | wake_up(&device->al_wait); | |
1070 | drbd_md_put_buffer(device); | |
1071 | drbd_resume_io(device); | |
b411b363 PR |
1072 | |
1073 | return rv; | |
1074 | } | |
1075 | ||
1076 | sector_t | |
b30ab791 | 1077 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1078 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1079 | { |
b30ab791 | 1080 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1081 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1082 | sector_t m_size; /* my size */ |
b411b363 PR |
1083 | sector_t size = 0; |
1084 | ||
1085 | m_size = drbd_get_max_capacity(bdev); | |
1086 | ||
b30ab791 | 1087 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1088 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1089 | p_size = m_size; |
1090 | } | |
1091 | ||
b411b363 PR |
1092 | if (p_size && m_size) { |
1093 | size = min_t(sector_t, p_size, m_size); | |
1094 | } else { | |
cccac985 LE |
1095 | if (la_size_sect) { |
1096 | size = la_size_sect; | |
b411b363 PR |
1097 | if (m_size && m_size < size) |
1098 | size = m_size; | |
1099 | if (p_size && p_size < size) | |
1100 | size = p_size; | |
1101 | } else { | |
1102 | if (m_size) | |
1103 | size = m_size; | |
1104 | if (p_size) | |
1105 | size = p_size; | |
1106 | } | |
1107 | } | |
1108 | ||
1109 | if (size == 0) | |
d0180171 | 1110 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1111 | |
1112 | if (u_size) { | |
1113 | if (u_size > size) | |
d0180171 | 1114 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1115 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1116 | else | |
1117 | size = u_size; | |
1118 | } | |
1119 | ||
1120 | return size; | |
1121 | } | |
1122 | ||
1123 | /** | |
1124 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1125 | * @device: DRBD device. |
b411b363 PR |
1126 | * |
1127 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1128 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1129 | * this function. | |
1130 | */ | |
b30ab791 | 1131 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1132 | { |
1133 | struct lru_cache *n, *t; | |
1134 | struct lc_element *e; | |
1135 | unsigned int in_use; | |
1136 | int i; | |
1137 | ||
b30ab791 AG |
1138 | if (device->act_log && |
1139 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1140 | return 0; |
1141 | ||
1142 | in_use = 0; | |
b30ab791 | 1143 | t = device->act_log; |
7ad651b5 | 1144 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1145 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1146 | |
1147 | if (n == NULL) { | |
d0180171 | 1148 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1149 | return -ENOMEM; |
1150 | } | |
b30ab791 | 1151 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1152 | if (t) { |
1153 | for (i = 0; i < t->nr_elements; i++) { | |
1154 | e = lc_element_by_index(t, i); | |
1155 | if (e->refcnt) | |
d0180171 | 1156 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1157 | e->lc_number, e->refcnt); |
1158 | in_use += e->refcnt; | |
1159 | } | |
1160 | } | |
1161 | if (!in_use) | |
b30ab791 AG |
1162 | device->act_log = n; |
1163 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1164 | if (in_use) { |
d0180171 | 1165 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1166 | lc_destroy(n); |
1167 | return -EBUSY; | |
1168 | } else { | |
d01efcee | 1169 | lc_destroy(t); |
b411b363 | 1170 | } |
b30ab791 | 1171 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1172 | return 0; |
1173 | } | |
1174 | ||
69ba1ee9 LE |
1175 | static void blk_queue_discard_granularity(struct request_queue *q, unsigned int granularity) |
1176 | { | |
1177 | q->limits.discard_granularity = granularity; | |
1178 | } | |
9104d31a LE |
1179 | |
1180 | static unsigned int drbd_max_discard_sectors(struct drbd_connection *connection) | |
1181 | { | |
1182 | /* when we introduced REQ_WRITE_SAME support, we also bumped | |
1183 | * our maximum supported batch bio size used for discards. */ | |
1184 | if (connection->agreed_features & DRBD_FF_WSAME) | |
1185 | return DRBD_MAX_BBIO_SECTORS; | |
1186 | /* before, with DRBD <= 8.4.6, we only allowed up to one AL_EXTENT_SIZE. */ | |
1187 | return AL_EXTENT_SIZE >> 9; | |
1188 | } | |
1189 | ||
69ba1ee9 LE |
1190 | static void decide_on_discard_support(struct drbd_device *device, |
1191 | struct request_queue *q, | |
1192 | struct request_queue *b, | |
1193 | bool discard_zeroes_if_aligned) | |
1194 | { | |
1195 | /* q = drbd device queue (device->rq_queue) | |
1196 | * b = backing device queue (device->ldev->backing_bdev->bd_disk->queue), | |
1197 | * or NULL if diskless | |
1198 | */ | |
1199 | struct drbd_connection *connection = first_peer_device(device)->connection; | |
1200 | bool can_do = b ? blk_queue_discard(b) : true; | |
1201 | ||
9104d31a | 1202 | if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_TRIM)) { |
69ba1ee9 LE |
1203 | can_do = false; |
1204 | drbd_info(connection, "peer DRBD too old, does not support TRIM: disabling discards\n"); | |
1205 | } | |
1206 | if (can_do) { | |
1207 | /* We don't care for the granularity, really. | |
1208 | * Stacking limits below should fix it for the local | |
1209 | * device. Whether or not it is a suitable granularity | |
1210 | * on the remote device is not our problem, really. If | |
1211 | * you care, you need to use devices with similar | |
1212 | * topology on all peers. */ | |
1213 | blk_queue_discard_granularity(q, 512); | |
9104d31a | 1214 | q->limits.max_discard_sectors = drbd_max_discard_sectors(connection); |
8b904b5b | 1215 | blk_queue_flag_set(QUEUE_FLAG_DISCARD, q); |
45c21793 | 1216 | q->limits.max_write_zeroes_sectors = drbd_max_discard_sectors(connection); |
69ba1ee9 | 1217 | } else { |
8b904b5b | 1218 | blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q); |
69ba1ee9 LE |
1219 | blk_queue_discard_granularity(q, 0); |
1220 | q->limits.max_discard_sectors = 0; | |
45c21793 | 1221 | q->limits.max_write_zeroes_sectors = 0; |
69ba1ee9 LE |
1222 | } |
1223 | } | |
1224 | ||
1225 | static void fixup_discard_if_not_supported(struct request_queue *q) | |
1226 | { | |
1227 | /* To avoid confusion, if this queue does not support discard, clear | |
1228 | * max_discard_sectors, which is what lsblk -D reports to the user. | |
1229 | * Older kernels got this wrong in "stack limits". | |
1230 | * */ | |
1231 | if (!blk_queue_discard(q)) { | |
1232 | blk_queue_max_discard_sectors(q, 0); | |
1233 | blk_queue_discard_granularity(q, 0); | |
1234 | } | |
1235 | } | |
1236 | ||
9104d31a LE |
1237 | static void decide_on_write_same_support(struct drbd_device *device, |
1238 | struct request_queue *q, | |
9de7e14a LE |
1239 | struct request_queue *b, struct o_qlim *o, |
1240 | bool disable_write_same) | |
9104d31a LE |
1241 | { |
1242 | struct drbd_peer_device *peer_device = first_peer_device(device); | |
1243 | struct drbd_connection *connection = peer_device->connection; | |
1244 | bool can_do = b ? b->limits.max_write_same_sectors : true; | |
1245 | ||
9de7e14a LE |
1246 | if (can_do && disable_write_same) { |
1247 | can_do = false; | |
1248 | drbd_info(peer_device, "WRITE_SAME disabled by config\n"); | |
1249 | } | |
1250 | ||
9104d31a LE |
1251 | if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_WSAME)) { |
1252 | can_do = false; | |
1253 | drbd_info(peer_device, "peer does not support WRITE_SAME\n"); | |
1254 | } | |
1255 | ||
1256 | if (o) { | |
1257 | /* logical block size; queue_logical_block_size(NULL) is 512 */ | |
1258 | unsigned int peer_lbs = be32_to_cpu(o->logical_block_size); | |
1259 | unsigned int me_lbs_b = queue_logical_block_size(b); | |
1260 | unsigned int me_lbs = queue_logical_block_size(q); | |
1261 | ||
1262 | if (me_lbs_b != me_lbs) { | |
1263 | drbd_warn(device, | |
1264 | "logical block size of local backend does not match (drbd:%u, backend:%u); was this a late attach?\n", | |
1265 | me_lbs, me_lbs_b); | |
1266 | /* rather disable write same than trigger some BUG_ON later in the scsi layer. */ | |
1267 | can_do = false; | |
1268 | } | |
1269 | if (me_lbs_b != peer_lbs) { | |
1270 | drbd_warn(peer_device, "logical block sizes do not match (me:%u, peer:%u); this may cause problems.\n", | |
1271 | me_lbs, peer_lbs); | |
1272 | if (can_do) { | |
1273 | drbd_dbg(peer_device, "logical block size mismatch: WRITE_SAME disabled.\n"); | |
1274 | can_do = false; | |
1275 | } | |
1276 | me_lbs = max(me_lbs, me_lbs_b); | |
1277 | /* We cannot change the logical block size of an in-use queue. | |
1278 | * We can only hope that access happens to be properly aligned. | |
1279 | * If not, the peer will likely produce an IO error, and detach. */ | |
1280 | if (peer_lbs > me_lbs) { | |
1281 | if (device->state.role != R_PRIMARY) { | |
1282 | blk_queue_logical_block_size(q, peer_lbs); | |
1283 | drbd_warn(peer_device, "logical block size set to %u\n", peer_lbs); | |
1284 | } else { | |
1285 | drbd_warn(peer_device, | |
1286 | "current Primary must NOT adjust logical block size (%u -> %u); hope for the best.\n", | |
1287 | me_lbs, peer_lbs); | |
1288 | } | |
1289 | } | |
1290 | } | |
1291 | if (can_do && !o->write_same_capable) { | |
1292 | /* If we introduce an open-coded write-same loop on the receiving side, | |
1293 | * the peer would present itself as "capable". */ | |
1294 | drbd_dbg(peer_device, "WRITE_SAME disabled (peer device not capable)\n"); | |
1295 | can_do = false; | |
1296 | } | |
1297 | } | |
1298 | ||
1299 | blk_queue_max_write_same_sectors(q, can_do ? DRBD_MAX_BBIO_SECTORS : 0); | |
1300 | } | |
1301 | ||
8fe39aac | 1302 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
9104d31a | 1303 | unsigned int max_bio_size, struct o_qlim *o) |
b411b363 | 1304 | { |
b30ab791 | 1305 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1306 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1307 | unsigned int max_segments = 0; | |
c1b3156f | 1308 | struct request_queue *b = NULL; |
dd4f699d LE |
1309 | struct disk_conf *dc; |
1310 | bool discard_zeroes_if_aligned = true; | |
9de7e14a | 1311 | bool disable_write_same = false; |
99432fcc | 1312 | |
8fe39aac PR |
1313 | if (bdev) { |
1314 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1315 | |
1316 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1317 | rcu_read_lock(); |
dd4f699d LE |
1318 | dc = rcu_dereference(device->ldev->disk_conf); |
1319 | max_segments = dc->max_bio_bvecs; | |
1320 | discard_zeroes_if_aligned = dc->discard_zeroes_if_aligned; | |
9de7e14a | 1321 | disable_write_same = dc->disable_write_same; |
daeda1cc | 1322 | rcu_read_unlock(); |
c1b3156f PR |
1323 | |
1324 | blk_set_stacking_limits(&q->limits); | |
99432fcc | 1325 | } |
b411b363 | 1326 | |
1816a2b4 LE |
1327 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1328 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1329 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
09cbfeaf | 1330 | blk_queue_segment_boundary(q, PAGE_SIZE-1); |
69ba1ee9 | 1331 | decide_on_discard_support(device, q, b, discard_zeroes_if_aligned); |
9de7e14a | 1332 | decide_on_write_same_support(device, q, b, o, disable_write_same); |
9104d31a | 1333 | |
c1b3156f | 1334 | if (b) { |
99432fcc PR |
1335 | blk_queue_stack_limits(q, b); |
1336 | ||
dc3b17cc JK |
1337 | if (q->backing_dev_info->ra_pages != |
1338 | b->backing_dev_info->ra_pages) { | |
d0180171 | 1339 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
dc3b17cc JK |
1340 | q->backing_dev_info->ra_pages, |
1341 | b->backing_dev_info->ra_pages); | |
1342 | q->backing_dev_info->ra_pages = | |
1343 | b->backing_dev_info->ra_pages; | |
99432fcc | 1344 | } |
99432fcc | 1345 | } |
69ba1ee9 | 1346 | fixup_discard_if_not_supported(q); |
99432fcc PR |
1347 | } |
1348 | ||
9104d31a | 1349 | void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_backing_dev *bdev, struct o_qlim *o) |
99432fcc | 1350 | { |
db141b2f | 1351 | unsigned int now, new, local, peer; |
99432fcc | 1352 | |
b30ab791 AG |
1353 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1354 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1355 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1356 | |
8fe39aac PR |
1357 | if (bdev) { |
1358 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1359 | device->local_max_bio_size = local; |
b411b363 | 1360 | } |
db141b2f | 1361 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1362 | |
1363 | /* We may ignore peer limits if the peer is modern enough. | |
1364 | Because new from 8.3.8 onwards the peer can use multiple | |
1365 | BIOs for a single peer_request */ | |
b30ab791 | 1366 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1367 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1368 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1369 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1370 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1371 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1372 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1373 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1374 | else | |
99432fcc | 1375 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1376 | |
fa090e70 LE |
1377 | /* We may later detach and re-attach on a disconnected Primary. |
1378 | * Avoid this setting to jump back in that case. | |
1379 | * We want to store what we know the peer DRBD can handle, | |
1380 | * not what the peer IO backend can handle. */ | |
1381 | if (peer > device->peer_max_bio_size) | |
1382 | device->peer_max_bio_size = peer; | |
1383 | } | |
db141b2f | 1384 | new = min(local, peer); |
99432fcc | 1385 | |
b30ab791 | 1386 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1387 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1388 | |
1389 | if (new != now) | |
d0180171 | 1390 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1391 | |
9104d31a | 1392 | drbd_setup_queue_param(device, bdev, new, o); |
b411b363 PR |
1393 | } |
1394 | ||
a18e9d1e | 1395 | /* Starts the worker thread */ |
bde89a9e | 1396 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1397 | { |
bde89a9e | 1398 | drbd_thread_start(&connection->worker); |
b5043c5e | 1399 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1400 | } |
1401 | ||
a18e9d1e | 1402 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1403 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1404 | { |
992d6e91 | 1405 | bool stop_threads; |
0500813f | 1406 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1407 | stop_threads = conn_all_vols_unconf(connection) && |
1408 | connection->cstate == C_STANDALONE; | |
0500813f | 1409 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 | 1410 | if (stop_threads) { |
668700b4 PR |
1411 | /* ack_receiver thread and ack_sender workqueue are implicitly |
1412 | * stopped by receiver in conn_disconnect() */ | |
bde89a9e AG |
1413 | drbd_thread_stop(&connection->receiver); |
1414 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1415 | } |
b411b363 PR |
1416 | } |
1417 | ||
0778286a | 1418 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1419 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1420 | { |
1421 | int s = 0; | |
1422 | ||
b30ab791 | 1423 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1424 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1425 | return; |
1426 | } | |
1427 | ||
b30ab791 | 1428 | drbd_al_shrink(device); |
0500813f | 1429 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1430 | if (device->state.conn < C_CONNECTED) |
1431 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1432 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1433 | lc_unlock(device->act_log); |
0778286a PR |
1434 | |
1435 | if (s) | |
d0180171 | 1436 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1437 | } |
1438 | ||
5979e361 LE |
1439 | |
1440 | static bool should_set_defaults(struct genl_info *info) | |
1441 | { | |
1442 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1443 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1444 | } | |
1445 | ||
5bbcf5e6 | 1446 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1447 | { |
5bbcf5e6 LE |
1448 | /* This is limited by 16 bit "slot" numbers, |
1449 | * and by available on-disk context storage. | |
1450 | * | |
1451 | * Also (u16)~0 is special (denotes a "free" extent). | |
1452 | * | |
1453 | * One transaction occupies one 4kB on-disk block, | |
1454 | * we have n such blocks in the on disk ring buffer, | |
1455 | * the "current" transaction may fail (n-1), | |
1456 | * and there is 919 slot numbers context information per transaction. | |
1457 | * | |
1458 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1459 | * so cap there first. | |
1460 | */ | |
1461 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1462 | const unsigned int sufficient_on_disk = | |
1463 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1464 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1465 | |
5bbcf5e6 LE |
1466 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1467 | ||
1468 | if (al_size_4k > sufficient_on_disk) | |
1469 | return max_al_nr; | |
1470 | ||
1471 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1472 | } |
1473 | ||
70df7092 LE |
1474 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1475 | { | |
1476 | return a->disk_barrier != b->disk_barrier || | |
1477 | a->disk_flushes != b->disk_flushes || | |
1478 | a->disk_drain != b->disk_drain; | |
1479 | } | |
1480 | ||
a5ca66c4 PR |
1481 | static void sanitize_disk_conf(struct drbd_device *device, struct disk_conf *disk_conf, |
1482 | struct drbd_backing_dev *nbc) | |
c5c23854 | 1483 | { |
a5ca66c4 PR |
1484 | struct request_queue * const q = nbc->backing_bdev->bd_disk->queue; |
1485 | ||
c5c23854 PR |
1486 | if (disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1487 | disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1488 | if (disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1489 | disk_conf->al_extents = drbd_al_extents_max(nbc); | |
a5ca66c4 | 1490 | |
48920ff2 | 1491 | if (!blk_queue_discard(q)) { |
65f5be35 LE |
1492 | if (disk_conf->rs_discard_granularity) { |
1493 | disk_conf->rs_discard_granularity = 0; /* disable feature */ | |
1494 | drbd_info(device, "rs_discard_granularity feature disabled\n"); | |
1495 | } | |
a5ca66c4 PR |
1496 | } |
1497 | ||
1498 | if (disk_conf->rs_discard_granularity) { | |
1499 | int orig_value = disk_conf->rs_discard_granularity; | |
1500 | int remainder; | |
1501 | ||
1502 | if (q->limits.discard_granularity > disk_conf->rs_discard_granularity) | |
1503 | disk_conf->rs_discard_granularity = q->limits.discard_granularity; | |
1504 | ||
1505 | remainder = disk_conf->rs_discard_granularity % q->limits.discard_granularity; | |
1506 | disk_conf->rs_discard_granularity += remainder; | |
1507 | ||
1508 | if (disk_conf->rs_discard_granularity > q->limits.max_discard_sectors << 9) | |
1509 | disk_conf->rs_discard_granularity = q->limits.max_discard_sectors << 9; | |
1510 | ||
1511 | if (disk_conf->rs_discard_granularity != orig_value) | |
1512 | drbd_info(device, "rs_discard_granularity changed to %d\n", | |
1513 | disk_conf->rs_discard_granularity); | |
1514 | } | |
c5c23854 PR |
1515 | } |
1516 | ||
f399002e LE |
1517 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1518 | { | |
a910b123 | 1519 | struct drbd_config_context adm_ctx; |
f399002e | 1520 | enum drbd_ret_code retcode; |
b30ab791 | 1521 | struct drbd_device *device; |
daeda1cc | 1522 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1523 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1524 | int err, fifo_size; |
f399002e | 1525 | |
a910b123 | 1526 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1527 | if (!adm_ctx.reply_skb) |
1528 | return retcode; | |
1529 | if (retcode != NO_ERROR) | |
9e276872 | 1530 | goto finish; |
f399002e | 1531 | |
b30ab791 | 1532 | device = adm_ctx.device; |
9e276872 | 1533 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1534 | |
1535 | /* we also need a disk | |
1536 | * to change the options on */ | |
b30ab791 | 1537 | if (!get_ldev(device)) { |
f399002e LE |
1538 | retcode = ERR_NO_DISK; |
1539 | goto out; | |
1540 | } | |
1541 | ||
daeda1cc | 1542 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1543 | if (!new_disk_conf) { |
f399002e LE |
1544 | retcode = ERR_NOMEM; |
1545 | goto fail; | |
1546 | } | |
1547 | ||
0500813f | 1548 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1549 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1550 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1551 | if (should_set_defaults(info)) |
b966b5dd | 1552 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1553 | |
5ecc72c3 | 1554 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1555 | if (err && err != -ENOMSG) { |
f399002e | 1556 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1557 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1558 | goto fail_unlock; |
f399002e LE |
1559 | } |
1560 | ||
5ecc72c3 LE |
1561 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1562 | new_disk_conf->resync_rate = 1; | |
f399002e | 1563 | |
a5ca66c4 | 1564 | sanitize_disk_conf(device, new_disk_conf, device->ldev); |
5bbcf5e6 LE |
1565 | |
1566 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1567 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1568 | |
5ecc72c3 | 1569 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1570 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1571 | new_plan = fifo_alloc(fifo_size); |
1572 | if (!new_plan) { | |
d0180171 | 1573 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1574 | retcode = ERR_NOMEM; |
daeda1cc | 1575 | goto fail_unlock; |
f399002e LE |
1576 | } |
1577 | } | |
1578 | ||
b30ab791 AG |
1579 | drbd_suspend_io(device); |
1580 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1581 | drbd_al_shrink(device); | |
1582 | err = drbd_check_al_size(device, new_disk_conf); | |
1583 | lc_unlock(device->act_log); | |
1584 | wake_up(&device->al_wait); | |
1585 | drbd_resume_io(device); | |
f399002e LE |
1586 | |
1587 | if (err) { | |
1588 | retcode = ERR_NOMEM; | |
daeda1cc | 1589 | goto fail_unlock; |
f399002e LE |
1590 | } |
1591 | ||
28bc3b8c | 1592 | lock_all_resources(); |
b30ab791 | 1593 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1594 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1595 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1596 | drbd_resync_after_changed(device); | |
dc97b708 | 1597 | } |
28bc3b8c | 1598 | unlock_all_resources(); |
f399002e | 1599 | |
daeda1cc PR |
1600 | if (retcode != NO_ERROR) |
1601 | goto fail_unlock; | |
f399002e | 1602 | |
813472ce | 1603 | if (new_plan) { |
b30ab791 AG |
1604 | old_plan = device->rs_plan_s; |
1605 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1606 | } |
9958c857 | 1607 | |
0500813f | 1608 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1609 | |
9a51ab1c | 1610 | if (new_disk_conf->al_updates) |
b30ab791 | 1611 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1612 | else |
b30ab791 | 1613 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1614 | |
691631c0 | 1615 | if (new_disk_conf->md_flushes) |
b30ab791 | 1616 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1617 | else |
b30ab791 | 1618 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1619 | |
70df7092 | 1620 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
f6ba8636 | 1621 | drbd_bump_write_ordering(device->resource, NULL, WO_BDEV_FLUSH); |
27eb13e9 | 1622 | |
9de7e14a LE |
1623 | if (old_disk_conf->discard_zeroes_if_aligned != new_disk_conf->discard_zeroes_if_aligned |
1624 | || old_disk_conf->disable_write_same != new_disk_conf->disable_write_same) | |
9104d31a | 1625 | drbd_reconsider_queue_parameters(device, device->ldev, NULL); |
dd4f699d | 1626 | |
b30ab791 | 1627 | drbd_md_sync(device); |
f399002e | 1628 | |
69a22773 AG |
1629 | if (device->state.conn >= C_CONNECTED) { |
1630 | struct drbd_peer_device *peer_device; | |
1631 | ||
1632 | for_each_peer_device(peer_device, device) | |
1633 | drbd_send_sync_param(peer_device); | |
1634 | } | |
f399002e | 1635 | |
daeda1cc PR |
1636 | synchronize_rcu(); |
1637 | kfree(old_disk_conf); | |
813472ce | 1638 | kfree(old_plan); |
b30ab791 | 1639 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1640 | goto success; |
1641 | ||
1642 | fail_unlock: | |
0500813f | 1643 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1644 | fail: |
5ecc72c3 | 1645 | kfree(new_disk_conf); |
813472ce | 1646 | kfree(new_plan); |
daeda1cc | 1647 | success: |
b30ab791 | 1648 | put_ldev(device); |
f399002e | 1649 | out: |
9e276872 LE |
1650 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1651 | finish: | |
a910b123 | 1652 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1653 | return 0; |
1654 | } | |
1655 | ||
63a7c8ad LE |
1656 | static struct block_device *open_backing_dev(struct drbd_device *device, |
1657 | const char *bdev_path, void *claim_ptr, bool do_bd_link) | |
1658 | { | |
1659 | struct block_device *bdev; | |
1660 | int err = 0; | |
1661 | ||
1662 | bdev = blkdev_get_by_path(bdev_path, | |
1663 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, claim_ptr); | |
1664 | if (IS_ERR(bdev)) { | |
1665 | drbd_err(device, "open(\"%s\") failed with %ld\n", | |
1666 | bdev_path, PTR_ERR(bdev)); | |
1667 | return bdev; | |
1668 | } | |
1669 | ||
1670 | if (!do_bd_link) | |
1671 | return bdev; | |
1672 | ||
1673 | err = bd_link_disk_holder(bdev, device->vdisk); | |
1674 | if (err) { | |
1675 | blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1676 | drbd_err(device, "bd_link_disk_holder(\"%s\", ...) failed with %d\n", | |
1677 | bdev_path, err); | |
1678 | bdev = ERR_PTR(err); | |
1679 | } | |
1680 | return bdev; | |
1681 | } | |
1682 | ||
1683 | static int open_backing_devices(struct drbd_device *device, | |
1684 | struct disk_conf *new_disk_conf, | |
1685 | struct drbd_backing_dev *nbc) | |
1686 | { | |
1687 | struct block_device *bdev; | |
1688 | ||
1689 | bdev = open_backing_dev(device, new_disk_conf->backing_dev, device, true); | |
1690 | if (IS_ERR(bdev)) | |
1691 | return ERR_OPEN_DISK; | |
1692 | nbc->backing_bdev = bdev; | |
1693 | ||
1694 | /* | |
1695 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1696 | * drbd sharing one meta device. TODO in that case, paranoia | |
1697 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1698 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1699 | * should check it for you already; but if you don't, or | |
1700 | * someone fooled it, we need to double check here) | |
1701 | */ | |
1702 | bdev = open_backing_dev(device, new_disk_conf->meta_dev, | |
1703 | /* claim ptr: device, if claimed exclusively; shared drbd_m_holder, | |
1704 | * if potentially shared with other drbd minors */ | |
1705 | (new_disk_conf->meta_dev_idx < 0) ? (void*)device : (void*)drbd_m_holder, | |
1706 | /* avoid double bd_claim_by_disk() for the same (source,target) tuple, | |
1707 | * as would happen with internal metadata. */ | |
1708 | (new_disk_conf->meta_dev_idx != DRBD_MD_INDEX_FLEX_INT && | |
1709 | new_disk_conf->meta_dev_idx != DRBD_MD_INDEX_INTERNAL)); | |
1710 | if (IS_ERR(bdev)) | |
1711 | return ERR_OPEN_MD_DISK; | |
1712 | nbc->md_bdev = bdev; | |
1713 | return NO_ERROR; | |
1714 | } | |
1715 | ||
1716 | static void close_backing_dev(struct drbd_device *device, struct block_device *bdev, | |
1717 | bool do_bd_unlink) | |
1718 | { | |
1719 | if (!bdev) | |
1720 | return; | |
1721 | if (do_bd_unlink) | |
1722 | bd_unlink_disk_holder(bdev, device->vdisk); | |
1723 | blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1724 | } | |
1725 | ||
1726 | void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev) | |
1727 | { | |
1728 | if (ldev == NULL) | |
1729 | return; | |
1730 | ||
1731 | close_backing_dev(device, ldev->md_bdev, ldev->md_bdev != ldev->backing_bdev); | |
1732 | close_backing_dev(device, ldev->backing_bdev, true); | |
1733 | ||
1734 | kfree(ldev->disk_conf); | |
1735 | kfree(ldev); | |
1736 | } | |
1737 | ||
3b98c0c2 | 1738 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1739 | { |
a910b123 | 1740 | struct drbd_config_context adm_ctx; |
b30ab791 | 1741 | struct drbd_device *device; |
44a4d551 LE |
1742 | struct drbd_peer_device *peer_device; |
1743 | struct drbd_connection *connection; | |
3b98c0c2 | 1744 | int err; |
116676ca | 1745 | enum drbd_ret_code retcode; |
b411b363 PR |
1746 | enum determine_dev_size dd; |
1747 | sector_t max_possible_sectors; | |
1748 | sector_t min_md_device_sectors; | |
1749 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1750 | struct disk_conf *new_disk_conf = NULL; |
b411b363 | 1751 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1752 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1753 | union drbd_state ns, os; |
f2024e7c | 1754 | enum drbd_state_rv rv; |
44ed167d | 1755 | struct net_conf *nc; |
b411b363 | 1756 | |
a910b123 | 1757 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1758 | if (!adm_ctx.reply_skb) |
1759 | return retcode; | |
1760 | if (retcode != NO_ERROR) | |
40cbf085 | 1761 | goto finish; |
b411b363 | 1762 | |
b30ab791 | 1763 | device = adm_ctx.device; |
9e276872 | 1764 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 | 1765 | peer_device = first_peer_device(device); |
3b8a44f8 | 1766 | connection = peer_device->connection; |
44a4d551 | 1767 | conn_reconfig_start(connection); |
b411b363 PR |
1768 | |
1769 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1770 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1771 | retcode = ERR_DISK_CONFIGURED; |
1772 | goto fail; | |
1773 | } | |
82f59cc6 LE |
1774 | /* It may just now have detached because of IO error. Make sure |
1775 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1776 | * e.g. if someone calls attach from the on-io-error handler, | |
1777 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1778 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1779 | |
383606e0 | 1780 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1781 | clear_bit(FORCE_DETACH, &device->flags); |
1782 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1783 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1784 | |
0029d624 | 1785 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1786 | device->rs_total = 0; |
1787 | device->rs_failed = 0; | |
1788 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1789 | |
3b98c0c2 | 1790 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1791 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1792 | if (!nbc) { | |
1793 | retcode = ERR_NOMEM; | |
1794 | goto fail; | |
1795 | } | |
9f2247bb PR |
1796 | spin_lock_init(&nbc->md.uuid_lock); |
1797 | ||
daeda1cc PR |
1798 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1799 | if (!new_disk_conf) { | |
1800 | retcode = ERR_NOMEM; | |
b411b363 PR |
1801 | goto fail; |
1802 | } | |
daeda1cc | 1803 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1804 | |
daeda1cc PR |
1805 | set_disk_conf_defaults(new_disk_conf); |
1806 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1807 | if (err) { |
b411b363 | 1808 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1809 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1810 | goto fail; |
1811 | } | |
1812 | ||
5bbcf5e6 LE |
1813 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1814 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1815 | |
9958c857 PR |
1816 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1817 | if (!new_plan) { | |
1818 | retcode = ERR_NOMEM; | |
1819 | goto fail; | |
1820 | } | |
1821 | ||
daeda1cc | 1822 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1823 | retcode = ERR_MD_IDX_INVALID; |
1824 | goto fail; | |
1825 | } | |
1826 | ||
44ed167d | 1827 | rcu_read_lock(); |
44a4d551 | 1828 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1829 | if (nc) { |
daeda1cc | 1830 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1831 | rcu_read_unlock(); |
47ff2d0a | 1832 | retcode = ERR_STONITH_AND_PROT_A; |
28bc3b8c | 1833 | goto fail; |
47ff2d0a PR |
1834 | } |
1835 | } | |
44ed167d | 1836 | rcu_read_unlock(); |
47ff2d0a | 1837 | |
63a7c8ad LE |
1838 | retcode = open_backing_devices(device, new_disk_conf, nbc); |
1839 | if (retcode != NO_ERROR) | |
28bc3b8c | 1840 | goto fail; |
b411b363 | 1841 | |
e525fd89 | 1842 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1843 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1844 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1845 | retcode = ERR_MD_IDX_INVALID; |
28bc3b8c | 1846 | goto fail; |
b411b363 PR |
1847 | } |
1848 | ||
1849 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1850 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1851 | offsetof(struct bm_extent, lce)); |
1852 | if (!resync_lru) { | |
1853 | retcode = ERR_NOMEM; | |
28bc3b8c | 1854 | goto fail; |
b411b363 PR |
1855 | } |
1856 | ||
c04ccaa6 LE |
1857 | /* Read our meta data super block early. |
1858 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1859 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 | 1860 | if (retcode != NO_ERROR) |
28bc3b8c | 1861 | goto fail; |
b411b363 | 1862 | |
a5ca66c4 | 1863 | sanitize_disk_conf(device, new_disk_conf, nbc); |
5bbcf5e6 | 1864 | |
daeda1cc | 1865 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1866 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1867 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1868 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1869 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1870 | goto fail; |
b411b363 PR |
1871 | } |
1872 | ||
daeda1cc | 1873 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1874 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1875 | /* at least one MB, otherwise it does not make sense */ | |
1876 | min_md_device_sectors = (2<<10); | |
1877 | } else { | |
1878 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1879 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1880 | } |
1881 | ||
b411b363 | 1882 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1883 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1884 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1885 | "at least %llu sectors needed for this meta-disk type\n", |
1886 | (unsigned long long) min_md_device_sectors); | |
28bc3b8c | 1887 | goto fail; |
b411b363 PR |
1888 | } |
1889 | ||
1890 | /* Make sure the new disk is big enough | |
1891 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1892 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1893 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1894 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1895 | goto fail; |
b411b363 PR |
1896 | } |
1897 | ||
1898 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1899 | ||
1352994b | 1900 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1901 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1902 | "to currently maximum possible %llu sectors <==\n", |
1903 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1904 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1905 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1906 | "meta data may help <<==\n"); |
1907 | } | |
1908 | ||
b30ab791 | 1909 | drbd_suspend_io(device); |
b411b363 | 1910 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1911 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1912 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1913 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1914 | * As barriers are counted per resource, | |
1915 | * we'd need to suspend io on all devices of a resource. | |
1916 | */ | |
b30ab791 | 1917 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1918 | /* and for any other previously queued work */ |
44a4d551 | 1919 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1920 | |
b30ab791 | 1921 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1922 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1923 | drbd_resume_io(device); |
f2024e7c | 1924 | if (rv < SS_SUCCESS) |
28bc3b8c | 1925 | goto fail; |
b411b363 | 1926 | |
b30ab791 | 1927 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1928 | goto force_diskless; |
1929 | ||
b30ab791 AG |
1930 | if (!device->bitmap) { |
1931 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1932 | retcode = ERR_NOMEM; |
1933 | goto force_diskless_dec; | |
1934 | } | |
1935 | } | |
1936 | ||
b30ab791 | 1937 | if (device->state.conn < C_CONNECTED && |
babea49e | 1938 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1939 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1940 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1941 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1942 | retcode = ERR_DATA_NOT_CURRENT; |
1943 | goto force_diskless_dec; | |
1944 | } | |
1945 | ||
1946 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1947 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1948 | retcode = ERR_NOMEM; |
1949 | goto force_diskless_dec; | |
1950 | } | |
1951 | ||
1952 | /* Prevent shrinking of consistent devices ! */ | |
1953 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1954 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1955 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1956 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1957 | goto force_diskless_dec; |
1958 | } | |
1959 | ||
28bc3b8c AG |
1960 | lock_all_resources(); |
1961 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); | |
1962 | if (retcode != NO_ERROR) { | |
1963 | unlock_all_resources(); | |
1964 | goto force_diskless_dec; | |
1965 | } | |
1966 | ||
b411b363 PR |
1967 | /* Reset the "barriers don't work" bits here, then force meta data to |
1968 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1969 | if (new_disk_conf->md_flushes) |
b30ab791 | 1970 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1971 | else |
b30ab791 | 1972 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1973 | |
1974 | /* Point of no return reached. | |
1975 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1976 | * now device takes over responsibility, and the state engine should |
b411b363 | 1977 | * clean it up somewhere. */ |
0b0ba1ef | 1978 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1979 | device->ldev = nbc; |
1980 | device->resync = resync_lru; | |
1981 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1982 | nbc = NULL; |
1983 | resync_lru = NULL; | |
daeda1cc | 1984 | new_disk_conf = NULL; |
9958c857 | 1985 | new_plan = NULL; |
b411b363 | 1986 | |
1ec317d3 | 1987 | drbd_resync_after_changed(device); |
f6ba8636 | 1988 | drbd_bump_write_ordering(device->resource, device->ldev, WO_BDEV_FLUSH); |
28bc3b8c | 1989 | unlock_all_resources(); |
b411b363 | 1990 | |
b30ab791 AG |
1991 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1992 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1993 | else |
b30ab791 | 1994 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1995 | |
b30ab791 | 1996 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1997 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1998 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1999 | |
b30ab791 AG |
2000 | device->send_cnt = 0; |
2001 | device->recv_cnt = 0; | |
2002 | device->read_cnt = 0; | |
2003 | device->writ_cnt = 0; | |
b411b363 | 2004 | |
9104d31a | 2005 | drbd_reconsider_queue_parameters(device, device->ldev, NULL); |
b411b363 PR |
2006 | |
2007 | /* If I am currently not R_PRIMARY, | |
2008 | * but meta data primary indicator is set, | |
2009 | * I just now recover from a hard crash, | |
2010 | * and have been R_PRIMARY before that crash. | |
2011 | * | |
2012 | * Now, if I had no connection before that crash | |
2013 | * (have been degraded R_PRIMARY), chances are that | |
2014 | * I won't find my peer now either. | |
2015 | * | |
2016 | * In that case, and _only_ in that case, | |
2017 | * we use the degr-wfc-timeout instead of the default, | |
2018 | * so we can automatically recover from a crash of a | |
2019 | * degraded but active "cluster" after a certain timeout. | |
2020 | */ | |
b30ab791 AG |
2021 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
2022 | if (device->state.role != R_PRIMARY && | |
2023 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
2024 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
2025 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 2026 | |
b30ab791 | 2027 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 2028 | if (dd <= DS_ERROR) { |
b411b363 PR |
2029 | retcode = ERR_NOMEM_BITMAP; |
2030 | goto force_diskless_dec; | |
e96c9633 | 2031 | } else if (dd == DS_GREW) |
b30ab791 | 2032 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 2033 | |
b30ab791 AG |
2034 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
2035 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
2036 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 2037 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 2038 | "(aka FullSync)\n"); |
b30ab791 | 2039 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 2040 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
2041 | retcode = ERR_IO_MD_DISK; |
2042 | goto force_diskless_dec; | |
2043 | } | |
2044 | } else { | |
b30ab791 | 2045 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 2046 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
2047 | retcode = ERR_IO_MD_DISK; |
2048 | goto force_diskless_dec; | |
2049 | } | |
b411b363 PR |
2050 | } |
2051 | ||
b30ab791 AG |
2052 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
2053 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 2054 | |
0500813f | 2055 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 2056 | os = drbd_read_state(device); |
78bae59b | 2057 | ns = os; |
b411b363 PR |
2058 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
2059 | otherwise investigate MDF_WasUpToDate... | |
2060 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
2061 | otherwise into D_CONSISTENT state. | |
2062 | */ | |
b30ab791 AG |
2063 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
2064 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
2065 | ns.disk = D_CONSISTENT; |
2066 | else | |
2067 | ns.disk = D_OUTDATED; | |
2068 | } else { | |
2069 | ns.disk = D_INCONSISTENT; | |
2070 | } | |
2071 | ||
b30ab791 | 2072 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
2073 | ns.pdsk = D_OUTDATED; |
2074 | ||
daeda1cc PR |
2075 | rcu_read_lock(); |
2076 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 2077 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
2078 | ns.disk = D_UP_TO_DATE; |
2079 | ||
2080 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
2081 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
2082 | this point, because drbd_request_state() modifies these | |
2083 | flags. */ | |
2084 | ||
b30ab791 AG |
2085 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
2086 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 2087 | else |
b30ab791 | 2088 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
2089 | |
2090 | rcu_read_unlock(); | |
2091 | ||
b411b363 PR |
2092 | /* In case we are C_CONNECTED postpone any decision on the new disk |
2093 | state after the negotiation phase. */ | |
b30ab791 AG |
2094 | if (device->state.conn == C_CONNECTED) { |
2095 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
2096 | ns.i = os.i; |
2097 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
2098 | |
2099 | /* We expect to receive up-to-date UUIDs soon. | |
2100 | To avoid a race in receive_state, free p_uuid while | |
2101 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
2102 | kfree(device->p_uuid); |
2103 | device->p_uuid = NULL; | |
b411b363 PR |
2104 | } |
2105 | ||
b30ab791 | 2106 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 2107 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
2108 | |
2109 | if (rv < SS_SUCCESS) | |
2110 | goto force_diskless_dec; | |
2111 | ||
b30ab791 | 2112 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 2113 | |
b30ab791 AG |
2114 | if (device->state.role == R_PRIMARY) |
2115 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 2116 | else |
b30ab791 | 2117 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 2118 | |
b30ab791 AG |
2119 | drbd_md_mark_dirty(device); |
2120 | drbd_md_sync(device); | |
b411b363 | 2121 | |
b30ab791 AG |
2122 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
2123 | put_ldev(device); | |
44a4d551 | 2124 | conn_reconfig_done(connection); |
9e276872 | 2125 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2126 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2127 | return 0; |
2128 | ||
2129 | force_diskless_dec: | |
b30ab791 | 2130 | put_ldev(device); |
b411b363 | 2131 | force_diskless: |
b30ab791 AG |
2132 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
2133 | drbd_md_sync(device); | |
b411b363 | 2134 | fail: |
44a4d551 | 2135 | conn_reconfig_done(connection); |
b411b363 | 2136 | if (nbc) { |
63a7c8ad LE |
2137 | close_backing_dev(device, nbc->md_bdev, nbc->md_bdev != nbc->backing_bdev); |
2138 | close_backing_dev(device, nbc->backing_bdev, true); | |
b411b363 PR |
2139 | kfree(nbc); |
2140 | } | |
daeda1cc | 2141 | kfree(new_disk_conf); |
b411b363 | 2142 | lc_destroy(resync_lru); |
9958c857 | 2143 | kfree(new_plan); |
9e276872 | 2144 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 2145 | finish: |
a910b123 | 2146 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2147 | return 0; |
2148 | } | |
2149 | ||
b30ab791 | 2150 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 2151 | { |
cdfda633 | 2152 | if (force) { |
b30ab791 AG |
2153 | set_bit(FORCE_DETACH, &device->flags); |
2154 | drbd_force_state(device, NS(disk, D_FAILED)); | |
33d32fa7 | 2155 | return SS_SUCCESS; |
02ee8f95 PR |
2156 | } |
2157 | ||
33d32fa7 | 2158 | return drbd_request_detach_interruptible(device); |
b411b363 PR |
2159 | } |
2160 | ||
82f59cc6 LE |
2161 | /* Detaching the disk is a process in multiple stages. First we need to lock |
2162 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
2163 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
2164 | * internal references as well. | |
2165 | * Only then we have finally detached. */ | |
3b98c0c2 | 2166 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2167 | { |
a910b123 | 2168 | struct drbd_config_context adm_ctx; |
116676ca | 2169 | enum drbd_ret_code retcode; |
cdfda633 PR |
2170 | struct detach_parms parms = { }; |
2171 | int err; | |
b411b363 | 2172 | |
a910b123 | 2173 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2174 | if (!adm_ctx.reply_skb) |
2175 | return retcode; | |
2176 | if (retcode != NO_ERROR) | |
2177 | goto out; | |
b411b363 | 2178 | |
cdfda633 PR |
2179 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
2180 | err = detach_parms_from_attrs(&parms, info); | |
2181 | if (err) { | |
2182 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2183 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
2184 | goto out; |
2185 | } | |
b411b363 PR |
2186 | } |
2187 | ||
9e276872 | 2188 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2189 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 2190 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2191 | out: |
a910b123 | 2192 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2193 | return 0; |
2194 | } | |
b411b363 | 2195 | |
bde89a9e | 2196 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 2197 | { |
c06ece6b | 2198 | struct drbd_peer_device *peer_device; |
695d08fa | 2199 | bool rv = false; |
f399002e LE |
2200 | int vnr; |
2201 | ||
695d08fa | 2202 | rcu_read_lock(); |
c06ece6b AG |
2203 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2204 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2205 | if (device->state.conn == C_SYNC_SOURCE || |
2206 | device->state.conn == C_SYNC_TARGET || | |
2207 | device->state.conn == C_PAUSED_SYNC_S || | |
2208 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
2209 | rv = true; |
2210 | break; | |
2211 | } | |
b411b363 | 2212 | } |
695d08fa | 2213 | rcu_read_unlock(); |
b411b363 | 2214 | |
695d08fa | 2215 | return rv; |
f399002e | 2216 | } |
47ff2d0a | 2217 | |
bde89a9e | 2218 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 2219 | { |
c06ece6b | 2220 | struct drbd_peer_device *peer_device; |
695d08fa | 2221 | bool rv = false; |
f399002e LE |
2222 | int vnr; |
2223 | ||
695d08fa | 2224 | rcu_read_lock(); |
c06ece6b AG |
2225 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2226 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2227 | if (device->state.conn == C_VERIFY_S || |
2228 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
2229 | rv = true; |
2230 | break; | |
47ff2d0a PR |
2231 | } |
2232 | } | |
695d08fa | 2233 | rcu_read_unlock(); |
b411b363 | 2234 | |
695d08fa | 2235 | return rv; |
f399002e | 2236 | } |
422028b1 | 2237 | |
cd64397c | 2238 | static enum drbd_ret_code |
270eb5c9 | 2239 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 2240 | { |
c06ece6b | 2241 | struct drbd_peer_device *peer_device; |
cd64397c | 2242 | int i; |
b411b363 | 2243 | |
270eb5c9 AG |
2244 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2245 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2246 | return ERR_NEED_APV_100; |
b411b363 | 2247 | |
270eb5c9 | 2248 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2249 | return ERR_NEED_APV_100; |
2250 | ||
270eb5c9 | 2251 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2252 | return ERR_NEED_APV_100; |
b411b363 PR |
2253 | } |
2254 | ||
270eb5c9 | 2255 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2256 | conn_highest_role(connection) == R_PRIMARY && |
2257 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2258 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2259 | |
270eb5c9 AG |
2260 | if (new_net_conf->two_primaries && |
2261 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2262 | return ERR_NOT_PROTO_C; |
2263 | ||
c06ece6b AG |
2264 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2265 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2266 | if (get_ldev(device)) { |
2267 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2268 | put_ldev(device); | |
270eb5c9 | 2269 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2270 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2271 | } |
270eb5c9 | 2272 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2273 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2274 | } |
2275 | ||
270eb5c9 | 2276 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2277 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2278 | |
cd64397c PR |
2279 | return NO_ERROR; |
2280 | } | |
b411b363 | 2281 | |
44ed167d | 2282 | static enum drbd_ret_code |
270eb5c9 | 2283 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d | 2284 | { |
e9d5d4a0 | 2285 | enum drbd_ret_code rv; |
c06ece6b | 2286 | struct drbd_peer_device *peer_device; |
44ed167d | 2287 | int i; |
b411b363 | 2288 | |
44ed167d | 2289 | rcu_read_lock(); |
270eb5c9 | 2290 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2291 | rcu_read_unlock(); |
b411b363 | 2292 | |
179e20b8 | 2293 | /* connection->peer_devices protected by genl_lock() here */ |
c06ece6b AG |
2294 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2295 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2296 | if (!device->bitmap) { |
2297 | if (drbd_bm_init(device)) | |
44ed167d | 2298 | return ERR_NOMEM; |
b411b363 PR |
2299 | } |
2300 | } | |
2301 | ||
44ed167d PR |
2302 | return rv; |
2303 | } | |
b411b363 | 2304 | |
0fd0ea06 | 2305 | struct crypto { |
9534d671 HX |
2306 | struct crypto_ahash *verify_tfm; |
2307 | struct crypto_ahash *csums_tfm; | |
2308 | struct crypto_shash *cram_hmac_tfm; | |
2309 | struct crypto_ahash *integrity_tfm; | |
0fd0ea06 | 2310 | }; |
b411b363 | 2311 | |
0fd0ea06 | 2312 | static int |
9534d671 | 2313 | alloc_shash(struct crypto_shash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2314 | { |
2315 | if (!tfm_name[0]) | |
2316 | return NO_ERROR; | |
b411b363 | 2317 | |
9534d671 HX |
2318 | *tfm = crypto_alloc_shash(tfm_name, 0, 0); |
2319 | if (IS_ERR(*tfm)) { | |
2320 | *tfm = NULL; | |
2321 | return err_alg; | |
2322 | } | |
2323 | ||
2324 | return NO_ERROR; | |
2325 | } | |
2326 | ||
2327 | static int | |
2328 | alloc_ahash(struct crypto_ahash **tfm, char *tfm_name, int err_alg) | |
2329 | { | |
2330 | if (!tfm_name[0]) | |
2331 | return NO_ERROR; | |
2332 | ||
2333 | *tfm = crypto_alloc_ahash(tfm_name, 0, CRYPTO_ALG_ASYNC); | |
0fd0ea06 PR |
2334 | if (IS_ERR(*tfm)) { |
2335 | *tfm = NULL; | |
2336 | return err_alg; | |
b411b363 | 2337 | } |
b411b363 | 2338 | |
0fd0ea06 PR |
2339 | return NO_ERROR; |
2340 | } | |
b411b363 | 2341 | |
0fd0ea06 | 2342 | static enum drbd_ret_code |
270eb5c9 | 2343 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2344 | { |
2345 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2346 | enum drbd_ret_code rv; | |
0fd0ea06 | 2347 | |
9534d671 HX |
2348 | rv = alloc_ahash(&crypto->csums_tfm, new_net_conf->csums_alg, |
2349 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
2350 | if (rv != NO_ERROR) |
2351 | return rv; | |
9534d671 HX |
2352 | rv = alloc_ahash(&crypto->verify_tfm, new_net_conf->verify_alg, |
2353 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
2354 | if (rv != NO_ERROR) |
2355 | return rv; | |
9534d671 HX |
2356 | rv = alloc_ahash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
2357 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
2358 | if (rv != NO_ERROR) |
2359 | return rv; | |
270eb5c9 | 2360 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2361 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2362 | new_net_conf->cram_hmac_alg); |
b411b363 | 2363 | |
9534d671 HX |
2364 | rv = alloc_shash(&crypto->cram_hmac_tfm, hmac_name, |
2365 | ERR_AUTH_ALG); | |
b411b363 PR |
2366 | } |
2367 | ||
0fd0ea06 PR |
2368 | return rv; |
2369 | } | |
b411b363 | 2370 | |
0fd0ea06 PR |
2371 | static void free_crypto(struct crypto *crypto) |
2372 | { | |
9534d671 HX |
2373 | crypto_free_shash(crypto->cram_hmac_tfm); |
2374 | crypto_free_ahash(crypto->integrity_tfm); | |
2375 | crypto_free_ahash(crypto->csums_tfm); | |
2376 | crypto_free_ahash(crypto->verify_tfm); | |
0fd0ea06 | 2377 | } |
b411b363 | 2378 | |
f399002e LE |
2379 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2380 | { | |
a910b123 | 2381 | struct drbd_config_context adm_ctx; |
f399002e | 2382 | enum drbd_ret_code retcode; |
bde89a9e | 2383 | struct drbd_connection *connection; |
270eb5c9 | 2384 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2385 | int err; |
2386 | int ovr; /* online verify running */ | |
2387 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2388 | struct crypto crypto = { }; |
b411b363 | 2389 | |
a910b123 | 2390 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2391 | if (!adm_ctx.reply_skb) |
2392 | return retcode; | |
2393 | if (retcode != NO_ERROR) | |
9e276872 | 2394 | goto finish; |
b411b363 | 2395 | |
bde89a9e | 2396 | connection = adm_ctx.connection; |
9e276872 | 2397 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2398 | |
270eb5c9 AG |
2399 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2400 | if (!new_net_conf) { | |
f399002e LE |
2401 | retcode = ERR_NOMEM; |
2402 | goto out; | |
2403 | } | |
b411b363 | 2404 | |
bde89a9e | 2405 | conn_reconfig_start(connection); |
b411b363 | 2406 | |
bde89a9e | 2407 | mutex_lock(&connection->data.mutex); |
0500813f | 2408 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2409 | old_net_conf = connection->net_conf; |
2561b9c1 | 2410 | |
270eb5c9 | 2411 | if (!old_net_conf) { |
a910b123 | 2412 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2413 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2414 | goto fail; |
2415 | } | |
2416 | ||
270eb5c9 | 2417 | *new_net_conf = *old_net_conf; |
5979e361 | 2418 | if (should_set_defaults(info)) |
270eb5c9 | 2419 | set_net_conf_defaults(new_net_conf); |
f399002e | 2420 | |
270eb5c9 | 2421 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2422 | if (err && err != -ENOMSG) { |
f399002e | 2423 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2424 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2425 | goto fail; |
2561b9c1 | 2426 | } |
b411b363 | 2427 | |
270eb5c9 | 2428 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2429 | if (retcode != NO_ERROR) |
2430 | goto fail; | |
b411b363 | 2431 | |
f399002e | 2432 | /* re-sync running */ |
bde89a9e | 2433 | rsr = conn_resync_running(connection); |
270eb5c9 | 2434 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2435 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2436 | goto fail; |
b411b363 PR |
2437 | } |
2438 | ||
f399002e | 2439 | /* online verify running */ |
bde89a9e | 2440 | ovr = conn_ov_running(connection); |
270eb5c9 | 2441 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2442 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2443 | goto fail; |
f399002e | 2444 | } |
b411b363 | 2445 | |
270eb5c9 | 2446 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2447 | if (retcode != NO_ERROR) |
b411b363 | 2448 | goto fail; |
f399002e | 2449 | |
270eb5c9 | 2450 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2451 | |
2452 | if (!rsr) { | |
9534d671 | 2453 | crypto_free_ahash(connection->csums_tfm); |
bde89a9e | 2454 | connection->csums_tfm = crypto.csums_tfm; |
0fd0ea06 | 2455 | crypto.csums_tfm = NULL; |
f399002e LE |
2456 | } |
2457 | if (!ovr) { | |
9534d671 | 2458 | crypto_free_ahash(connection->verify_tfm); |
bde89a9e | 2459 | connection->verify_tfm = crypto.verify_tfm; |
0fd0ea06 | 2460 | crypto.verify_tfm = NULL; |
b411b363 PR |
2461 | } |
2462 | ||
9534d671 | 2463 | crypto_free_ahash(connection->integrity_tfm); |
bde89a9e AG |
2464 | connection->integrity_tfm = crypto.integrity_tfm; |
2465 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2466 | /* Do this without trying to take connection->data.mutex again. */ | |
2467 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2468 | |
9534d671 | 2469 | crypto_free_shash(connection->cram_hmac_tfm); |
bde89a9e | 2470 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; |
0fd0ea06 | 2471 | |
0500813f | 2472 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2473 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2474 | synchronize_rcu(); |
270eb5c9 | 2475 | kfree(old_net_conf); |
91fd4dad | 2476 | |
69a22773 AG |
2477 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2478 | struct drbd_peer_device *peer_device; | |
2479 | int vnr; | |
2480 | ||
2481 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2482 | drbd_send_sync_param(peer_device); | |
2483 | } | |
f399002e | 2484 | |
91fd4dad PR |
2485 | goto done; |
2486 | ||
b411b363 | 2487 | fail: |
0500813f | 2488 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2489 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2490 | free_crypto(&crypto); |
270eb5c9 | 2491 | kfree(new_net_conf); |
91fd4dad | 2492 | done: |
bde89a9e | 2493 | conn_reconfig_done(connection); |
f399002e | 2494 | out: |
9e276872 LE |
2495 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2496 | finish: | |
a910b123 | 2497 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2498 | return 0; |
2499 | } | |
2500 | ||
a2972846 AG |
2501 | static void connection_to_info(struct connection_info *info, |
2502 | struct drbd_connection *connection) | |
2503 | { | |
2504 | info->conn_connection_state = connection->cstate; | |
2505 | info->conn_role = conn_highest_peer(connection); | |
2506 | } | |
2507 | ||
2508 | static void peer_device_to_info(struct peer_device_info *info, | |
2509 | struct drbd_peer_device *peer_device) | |
2510 | { | |
2511 | struct drbd_device *device = peer_device->device; | |
2512 | ||
2513 | info->peer_repl_state = | |
2514 | max_t(enum drbd_conns, C_WF_REPORT_PARAMS, device->state.conn); | |
2515 | info->peer_disk_state = device->state.pdsk; | |
2516 | info->peer_resync_susp_user = device->state.user_isp; | |
2517 | info->peer_resync_susp_peer = device->state.peer_isp; | |
2518 | info->peer_resync_susp_dependency = device->state.aftr_isp; | |
2519 | } | |
2520 | ||
3b98c0c2 | 2521 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2522 | { |
a2972846 AG |
2523 | struct connection_info connection_info; |
2524 | enum drbd_notification_type flags; | |
2525 | unsigned int peer_devices = 0; | |
a910b123 | 2526 | struct drbd_config_context adm_ctx; |
c06ece6b | 2527 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2528 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2529 | struct crypto crypto = { }; |
77c556f6 | 2530 | struct drbd_resource *resource; |
bde89a9e | 2531 | struct drbd_connection *connection; |
3b98c0c2 LE |
2532 | enum drbd_ret_code retcode; |
2533 | int i; | |
2534 | int err; | |
b411b363 | 2535 | |
a910b123 | 2536 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2537 | |
3b98c0c2 LE |
2538 | if (!adm_ctx.reply_skb) |
2539 | return retcode; | |
2540 | if (retcode != NO_ERROR) | |
2541 | goto out; | |
089c075d | 2542 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2543 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2544 | retcode = ERR_INVALID_REQUEST; |
2545 | goto out; | |
2546 | } | |
b411b363 | 2547 | |
089c075d AG |
2548 | /* No need for _rcu here. All reconfiguration is |
2549 | * strictly serialized on genl_lock(). We are protected against | |
2550 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2551 | for_each_resource(resource, &drbd_resources) { |
2552 | for_each_connection(connection, resource) { | |
2553 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2554 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2555 | connection->my_addr_len)) { | |
2556 | retcode = ERR_LOCAL_ADDR; | |
2557 | goto out; | |
2558 | } | |
b411b363 | 2559 | |
77c556f6 AG |
2560 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2561 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2562 | connection->peer_addr_len)) { | |
2563 | retcode = ERR_PEER_ADDR; | |
2564 | goto out; | |
2565 | } | |
089c075d | 2566 | } |
b411b363 PR |
2567 | } |
2568 | ||
9e276872 | 2569 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2570 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2571 | conn_reconfig_start(connection); |
b411b363 | 2572 | |
bde89a9e | 2573 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2574 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2575 | goto fail; |
2576 | } | |
2577 | ||
a209b4ae | 2578 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2579 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2580 | if (!new_net_conf) { | |
b411b363 | 2581 | retcode = ERR_NOMEM; |
b411b363 PR |
2582 | goto fail; |
2583 | } | |
2584 | ||
270eb5c9 | 2585 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2586 | |
270eb5c9 | 2587 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2588 | if (err && err != -ENOMSG) { |
b411b363 | 2589 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2590 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2591 | goto fail; |
2592 | } | |
2593 | ||
270eb5c9 | 2594 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2595 | if (retcode != NO_ERROR) |
422028b1 | 2596 | goto fail; |
b411b363 | 2597 | |
270eb5c9 | 2598 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2599 | if (retcode != NO_ERROR) |
2600 | goto fail; | |
b411b363 | 2601 | |
270eb5c9 | 2602 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2603 | |
b5043c5e | 2604 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2605 | |
0500813f | 2606 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2607 | old_net_conf = connection->net_conf; |
2608 | if (old_net_conf) { | |
b411b363 | 2609 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2610 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2611 | goto fail; |
2612 | } | |
270eb5c9 | 2613 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2614 | |
bde89a9e AG |
2615 | conn_free_crypto(connection); |
2616 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2617 | connection->integrity_tfm = crypto.integrity_tfm; | |
2618 | connection->csums_tfm = crypto.csums_tfm; | |
2619 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2620 | |
bde89a9e AG |
2621 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2622 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2623 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2624 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2625 | |
a2972846 AG |
2626 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2627 | peer_devices++; | |
2628 | } | |
2629 | ||
2630 | connection_to_info(&connection_info, connection); | |
2631 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2632 | mutex_lock(¬ification_mutex); | |
2633 | notify_connection_state(NULL, 0, connection, &connection_info, NOTIFY_CREATE | flags); | |
2634 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
2635 | struct peer_device_info peer_device_info; | |
2636 | ||
2637 | peer_device_to_info(&peer_device_info, peer_device); | |
2638 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2639 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, NOTIFY_CREATE | flags); | |
2640 | } | |
2641 | mutex_unlock(¬ification_mutex); | |
0500813f | 2642 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2643 | |
695d08fa | 2644 | rcu_read_lock(); |
c06ece6b AG |
2645 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2646 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2647 | device->send_cnt = 0; |
2648 | device->recv_cnt = 0; | |
b411b363 | 2649 | } |
695d08fa | 2650 | rcu_read_unlock(); |
b411b363 | 2651 | |
bde89a9e | 2652 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2653 | |
bde89a9e | 2654 | conn_reconfig_done(connection); |
9e276872 | 2655 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2656 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2657 | return 0; |
b411b363 | 2658 | |
b411b363 | 2659 | fail: |
0fd0ea06 | 2660 | free_crypto(&crypto); |
270eb5c9 | 2661 | kfree(new_net_conf); |
b411b363 | 2662 | |
bde89a9e | 2663 | conn_reconfig_done(connection); |
9e276872 | 2664 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2665 | out: |
a910b123 | 2666 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2667 | return 0; |
2668 | } | |
2669 | ||
bde89a9e | 2670 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2671 | { |
2672 | enum drbd_state_rv rv; | |
85f75dd7 | 2673 | |
bde89a9e | 2674 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2675 | force ? CS_HARD : 0); |
85f75dd7 LE |
2676 | |
2677 | switch (rv) { | |
2678 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2679 | break; |
85f75dd7 LE |
2680 | case SS_ALREADY_STANDALONE: |
2681 | return SS_SUCCESS; | |
2682 | case SS_PRIMARY_NOP: | |
2683 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2684 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2685 | |
2686 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2687 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2688 | |
85f75dd7 LE |
2689 | break; |
2690 | case SS_CW_FAILED_BY_PEER: | |
2691 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2692 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2693 | disk, D_OUTDATED), 0); |
2694 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2695 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2696 | CS_HARD); |
b411b363 | 2697 | } |
85f75dd7 LE |
2698 | break; |
2699 | default:; | |
2700 | /* no special handling necessary */ | |
2701 | } | |
2702 | ||
f3dfa40a LE |
2703 | if (rv >= SS_SUCCESS) { |
2704 | enum drbd_state_rv rv2; | |
2705 | /* No one else can reconfigure the network while I am here. | |
2706 | * The state handling only uses drbd_thread_stop_nowait(), | |
2707 | * we want to really wait here until the receiver is no more. | |
2708 | */ | |
9693da23 | 2709 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2710 | |
2711 | /* Race breaker. This additional state change request may be | |
2712 | * necessary, if this was a forced disconnect during a receiver | |
2713 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2714 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2715 | * C_STANDALONE already, now, and this becomes a no-op. |
2716 | */ | |
bde89a9e | 2717 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2718 | CS_VERBOSE | CS_HARD); |
2719 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2720 | drbd_err(connection, |
f3dfa40a LE |
2721 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2722 | rv2); | |
a2972846 AG |
2723 | /* Unlike in DRBD 9, the state engine has generated |
2724 | * NOTIFY_DESTROY events before clearing connection->net_conf. */ | |
b411b363 | 2725 | } |
85f75dd7 LE |
2726 | return rv; |
2727 | } | |
b411b363 | 2728 | |
3b98c0c2 | 2729 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2730 | { |
a910b123 | 2731 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2732 | struct disconnect_parms parms; |
bde89a9e | 2733 | struct drbd_connection *connection; |
85f75dd7 | 2734 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2735 | enum drbd_ret_code retcode; |
2736 | int err; | |
2561b9c1 | 2737 | |
a910b123 | 2738 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2739 | if (!adm_ctx.reply_skb) |
2740 | return retcode; | |
2741 | if (retcode != NO_ERROR) | |
2561b9c1 | 2742 | goto fail; |
b411b363 | 2743 | |
bde89a9e | 2744 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2745 | memset(&parms, 0, sizeof(parms)); |
2746 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2747 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2748 | if (err) { |
2749 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2750 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2751 | goto fail; |
2752 | } | |
2753 | } | |
2754 | ||
9e276872 | 2755 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2756 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2757 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2758 | retcode = rv; /* FIXME: Type mismatch. */ |
2759 | else | |
2760 | retcode = NO_ERROR; | |
9e276872 | 2761 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2762 | fail: |
a910b123 | 2763 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2764 | return 0; |
2765 | } | |
2766 | ||
b30ab791 | 2767 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2768 | { |
2769 | int iass; /* I am sync source */ | |
2770 | ||
d0180171 | 2771 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2772 | if (device->state.role != device->state.peer) |
2773 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2774 | else |
a6b32bc3 | 2775 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2776 | |
2777 | if (iass) | |
b30ab791 | 2778 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2779 | else |
b30ab791 | 2780 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2781 | } |
2782 | ||
3b98c0c2 | 2783 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2784 | { |
a910b123 | 2785 | struct drbd_config_context adm_ctx; |
daeda1cc | 2786 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2787 | struct resize_parms rs; |
b30ab791 | 2788 | struct drbd_device *device; |
3b98c0c2 | 2789 | enum drbd_ret_code retcode; |
b411b363 | 2790 | enum determine_dev_size dd; |
d752b269 | 2791 | bool change_al_layout = false; |
6495d2c6 | 2792 | enum dds_flags ddsf; |
daeda1cc | 2793 | sector_t u_size; |
3b98c0c2 | 2794 | int err; |
b411b363 | 2795 | |
a910b123 | 2796 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2797 | if (!adm_ctx.reply_skb) |
2798 | return retcode; | |
2799 | if (retcode != NO_ERROR) | |
9e276872 | 2800 | goto finish; |
3b98c0c2 | 2801 | |
9e276872 | 2802 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2803 | device = adm_ctx.device; |
2804 | if (!get_ldev(device)) { | |
d752b269 PR |
2805 | retcode = ERR_NO_DISK; |
2806 | goto fail; | |
2807 | } | |
2808 | ||
3b98c0c2 | 2809 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2810 | rs.al_stripes = device->ldev->md.al_stripes; |
2811 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2812 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2813 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2814 | if (err) { |
3b98c0c2 | 2815 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2816 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2817 | goto fail_ldev; |
b411b363 PR |
2818 | } |
2819 | } | |
2820 | ||
b30ab791 | 2821 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2822 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2823 | goto fail_ldev; |
b411b363 | 2824 | } |
b411b363 | 2825 | |
b30ab791 AG |
2826 | if (device->state.role == R_SECONDARY && |
2827 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2828 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2829 | goto fail_ldev; |
b411b363 | 2830 | } |
b411b363 | 2831 | |
a6b32bc3 | 2832 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2833 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2834 | goto fail_ldev; |
6495d2c6 PR |
2835 | } |
2836 | ||
daeda1cc | 2837 | rcu_read_lock(); |
b30ab791 | 2838 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2839 | rcu_read_unlock(); |
2840 | if (u_size != (sector_t)rs.resize_size) { | |
2841 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2842 | if (!new_disk_conf) { | |
778f271d | 2843 | retcode = ERR_NOMEM; |
9bcd2521 | 2844 | goto fail_ldev; |
778f271d PR |
2845 | } |
2846 | } | |
2847 | ||
b30ab791 AG |
2848 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2849 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2850 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2851 | ||
2852 | if (al_size_k > (16 * 1024 * 1024)) { | |
2853 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2854 | goto fail_ldev; | |
2855 | } | |
2856 | ||
2857 | if (al_size_k < MD_32kB_SECT/2) { | |
2858 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2859 | goto fail_ldev; | |
2860 | } | |
2861 | ||
cdc6af8d | 2862 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2863 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2864 | goto fail_ldev; | |
2865 | } | |
2866 | ||
2867 | change_al_layout = true; | |
2868 | } | |
2869 | ||
b30ab791 AG |
2870 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2871 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2872 | |
daeda1cc | 2873 | if (new_disk_conf) { |
0500813f | 2874 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2875 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2876 | *new_disk_conf = *old_disk_conf; |
2877 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2878 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2879 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2880 | synchronize_rcu(); |
2881 | kfree(old_disk_conf); | |
70644786 | 2882 | new_disk_conf = NULL; |
b411b363 PR |
2883 | } |
2884 | ||
6495d2c6 | 2885 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2886 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2887 | drbd_md_sync(device); | |
2888 | put_ldev(device); | |
e96c9633 | 2889 | if (dd == DS_ERROR) { |
b411b363 PR |
2890 | retcode = ERR_NOMEM_BITMAP; |
2891 | goto fail; | |
d752b269 PR |
2892 | } else if (dd == DS_ERROR_SPACE_MD) { |
2893 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2894 | goto fail; | |
2895 | } else if (dd == DS_ERROR_SHRINK) { | |
2896 | retcode = ERR_IMPLICIT_SHRINK; | |
2897 | goto fail; | |
b411b363 | 2898 | } |
778f271d | 2899 | |
b30ab791 | 2900 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2901 | if (dd == DS_GREW) |
b30ab791 | 2902 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2903 | |
69a22773 AG |
2904 | drbd_send_uuids(first_peer_device(device)); |
2905 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2906 | } |
2907 | ||
b411b363 | 2908 | fail: |
9e276872 LE |
2909 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2910 | finish: | |
a910b123 | 2911 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2912 | return 0; |
b411b363 | 2913 | |
9bcd2521 | 2914 | fail_ldev: |
b30ab791 | 2915 | put_ldev(device); |
70644786 | 2916 | kfree(new_disk_conf); |
9bcd2521 | 2917 | goto fail; |
b411b363 | 2918 | } |
b411b363 | 2919 | |
f399002e | 2920 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2921 | { |
a910b123 | 2922 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2923 | enum drbd_ret_code retcode; |
b57a1e27 | 2924 | struct res_opts res_opts; |
f399002e | 2925 | int err; |
b411b363 | 2926 | |
a910b123 | 2927 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2928 | if (!adm_ctx.reply_skb) |
2929 | return retcode; | |
2930 | if (retcode != NO_ERROR) | |
2931 | goto fail; | |
b411b363 | 2932 | |
eb6bea67 | 2933 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2934 | if (should_set_defaults(info)) |
b966b5dd | 2935 | set_res_opts_defaults(&res_opts); |
b411b363 | 2936 | |
b57a1e27 | 2937 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2938 | if (err && err != -ENOMSG) { |
b411b363 | 2939 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2940 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2941 | goto fail; |
2942 | } | |
2943 | ||
9e276872 | 2944 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2945 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2946 | if (err) { |
2947 | retcode = ERR_INVALID_REQUEST; | |
2948 | if (err == -ENOMEM) | |
2949 | retcode = ERR_NOMEM; | |
b411b363 | 2950 | } |
9e276872 | 2951 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2952 | |
b411b363 | 2953 | fail: |
a910b123 | 2954 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2955 | return 0; |
2956 | } | |
2957 | ||
3b98c0c2 | 2958 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2959 | { |
a910b123 | 2960 | struct drbd_config_context adm_ctx; |
b30ab791 | 2961 | struct drbd_device *device; |
3b98c0c2 LE |
2962 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2963 | ||
a910b123 | 2964 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2965 | if (!adm_ctx.reply_skb) |
2966 | return retcode; | |
2967 | if (retcode != NO_ERROR) | |
2968 | goto out; | |
2969 | ||
b30ab791 | 2970 | device = adm_ctx.device; |
8fe39aac PR |
2971 | if (!get_ldev(device)) { |
2972 | retcode = ERR_NO_DISK; | |
2973 | goto out; | |
2974 | } | |
2975 | ||
2976 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2977 | |
194bfb32 | 2978 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2979 | * resync just being finished, wait for it before requesting a new resync. |
2980 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2981 | drbd_suspend_io(device); |
2982 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2983 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2984 | |
0b2dafcd PR |
2985 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2986 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2987 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2988 | */ |
b30ab791 AG |
2989 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2990 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2991 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2992 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2993 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2994 | retcode = ERR_IO_MD_DISK; | |
2995 | } | |
2996 | } else | |
b30ab791 AG |
2997 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2998 | drbd_resume_io(device); | |
9e276872 | 2999 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 3000 | put_ldev(device); |
3b98c0c2 | 3001 | out: |
a910b123 | 3002 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3003 | return 0; |
3004 | } | |
3005 | ||
3b98c0c2 LE |
3006 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
3007 | union drbd_state mask, union drbd_state val) | |
b411b363 | 3008 | { |
a910b123 | 3009 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3010 | enum drbd_ret_code retcode; |
194bfb32 | 3011 | |
a910b123 | 3012 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3013 | if (!adm_ctx.reply_skb) |
3014 | return retcode; | |
3015 | if (retcode != NO_ERROR) | |
3016 | goto out; | |
b411b363 | 3017 | |
9e276872 | 3018 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3019 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 3020 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3021 | out: |
a910b123 | 3022 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3023 | return 0; |
3024 | } | |
3025 | ||
8fe39aac | 3026 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
3027 | { |
3028 | int rv; | |
3029 | ||
b30ab791 AG |
3030 | rv = drbd_bmio_set_n_write(device); |
3031 | drbd_suspend_al(device); | |
0778286a PR |
3032 | return rv; |
3033 | } | |
3034 | ||
3b98c0c2 | 3035 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3036 | { |
a910b123 | 3037 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 3038 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 3039 | struct drbd_device *device; |
25b0d6c8 | 3040 | |
a910b123 | 3041 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
3042 | if (!adm_ctx.reply_skb) |
3043 | return retcode; | |
3044 | if (retcode != NO_ERROR) | |
3045 | goto out; | |
3046 | ||
b30ab791 | 3047 | device = adm_ctx.device; |
8fe39aac PR |
3048 | if (!get_ldev(device)) { |
3049 | retcode = ERR_NO_DISK; | |
3050 | goto out; | |
3051 | } | |
3052 | ||
3053 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 3054 | |
194bfb32 | 3055 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
3056 | * resync just being finished, wait for it before requesting a new resync. |
3057 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
3058 | drbd_suspend_io(device); |
3059 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 3060 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 3061 | |
0b2dafcd PR |
3062 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
3063 | * in the bitmap. Otherwise, try to start a resync handshake | |
3064 | * as sync source for full sync. | |
3065 | */ | |
b30ab791 | 3066 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
3067 | /* The peer will get a resync upon connect anyways. Just make that |
3068 | into a full resync. */ | |
b30ab791 | 3069 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 3070 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 3071 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
3072 | "set_n_write from invalidate_peer", |
3073 | BM_LOCKED_SET_ALLOWED)) | |
3074 | retcode = ERR_IO_MD_DISK; | |
3075 | } | |
3076 | } else | |
b30ab791 AG |
3077 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
3078 | drbd_resume_io(device); | |
9e276872 | 3079 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 3080 | put_ldev(device); |
25b0d6c8 | 3081 | out: |
a910b123 | 3082 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3083 | return 0; |
3084 | } | |
3085 | ||
3b98c0c2 | 3086 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3087 | { |
a910b123 | 3088 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3089 | enum drbd_ret_code retcode; |
b411b363 | 3090 | |
a910b123 | 3091 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3092 | if (!adm_ctx.reply_skb) |
3093 | return retcode; | |
3094 | if (retcode != NO_ERROR) | |
3095 | goto out; | |
b411b363 | 3096 | |
9e276872 | 3097 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3098 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 3099 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 3100 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3101 | out: |
a910b123 | 3102 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3103 | return 0; |
3104 | } | |
3105 | ||
3b98c0c2 | 3106 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3107 | { |
a910b123 | 3108 | struct drbd_config_context adm_ctx; |
da9fbc27 | 3109 | union drbd_dev_state s; |
3b98c0c2 LE |
3110 | enum drbd_ret_code retcode; |
3111 | ||
a910b123 | 3112 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3113 | if (!adm_ctx.reply_skb) |
3114 | return retcode; | |
3115 | if (retcode != NO_ERROR) | |
3116 | goto out; | |
b411b363 | 3117 | |
9e276872 | 3118 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
3119 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
3120 | s = adm_ctx.device->state; | |
cd88d030 PR |
3121 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
3122 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
3123 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
3124 | } else { | |
3125 | retcode = ERR_PAUSE_IS_CLEAR; | |
3126 | } | |
3127 | } | |
9e276872 | 3128 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3129 | out: |
a910b123 | 3130 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3131 | return 0; |
3132 | } | |
3133 | ||
3b98c0c2 | 3134 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3135 | { |
3b98c0c2 | 3136 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
3137 | } |
3138 | ||
3b98c0c2 | 3139 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3140 | { |
a910b123 | 3141 | struct drbd_config_context adm_ctx; |
b30ab791 | 3142 | struct drbd_device *device; |
3b98c0c2 LE |
3143 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
3144 | ||
a910b123 | 3145 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3146 | if (!adm_ctx.reply_skb) |
3147 | return retcode; | |
3148 | if (retcode != NO_ERROR) | |
3149 | goto out; | |
3150 | ||
9e276872 | 3151 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
3152 | device = adm_ctx.device; |
3153 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
9fa48269 LE |
3154 | if (get_ldev_if_state(device, D_ATTACHING)) { |
3155 | drbd_uuid_new_current(device); | |
3156 | put_ldev(device); | |
3157 | } else { | |
3158 | /* This is effectively a multi-stage "forced down". | |
3159 | * The NEW_CUR_UUID bit is supposedly only set, if we | |
3160 | * lost the replication connection, and are configured | |
3161 | * to freeze IO and wait for some fence-peer handler. | |
3162 | * So we still don't have a replication connection. | |
3163 | * And now we don't have a local disk either. After | |
3164 | * resume, we will fail all pending and new IO, because | |
3165 | * we don't have any data anymore. Which means we will | |
3166 | * eventually be able to terminate all users of this | |
3167 | * device, and then take it down. By bumping the | |
3168 | * "effective" data uuid, we make sure that you really | |
3169 | * need to tear down before you reconfigure, we will | |
3170 | * the refuse to re-connect or re-attach (because no | |
3171 | * matching real data uuid exists). | |
3172 | */ | |
3173 | u64 val; | |
3174 | get_random_bytes(&val, sizeof(u64)); | |
3175 | drbd_set_ed_uuid(device, val); | |
3176 | drbd_warn(device, "Resumed without access to data; please tear down before attempting to re-configure.\n"); | |
3177 | } | |
b30ab791 | 3178 | clear_bit(NEW_CUR_UUID, &device->flags); |
43a5182c | 3179 | } |
b30ab791 AG |
3180 | drbd_suspend_io(device); |
3181 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 3182 | if (retcode == SS_SUCCESS) { |
b30ab791 | 3183 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 3184 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 3185 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 3186 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 3187 | } |
b30ab791 | 3188 | drbd_resume_io(device); |
9e276872 | 3189 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3190 | out: |
a910b123 | 3191 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3192 | return 0; |
3193 | } | |
3194 | ||
3b98c0c2 | 3195 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3196 | { |
3b98c0c2 | 3197 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
3198 | } |
3199 | ||
251b8f8e AG |
3200 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
3201 | struct drbd_resource *resource, | |
3202 | struct drbd_connection *connection, | |
3203 | struct drbd_device *device) | |
b411b363 | 3204 | { |
543cc10b LE |
3205 | struct nlattr *nla; |
3206 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
3207 | if (!nla) | |
3208 | goto nla_put_failure; | |
251b8f8e AG |
3209 | if (device && |
3210 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 3211 | goto nla_put_failure; |
f597f6b8 | 3212 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 3213 | goto nla_put_failure; |
251b8f8e AG |
3214 | if (connection) { |
3215 | if (connection->my_addr_len && | |
3216 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
3217 | goto nla_put_failure; | |
3218 | if (connection->peer_addr_len && | |
3219 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
3220 | goto nla_put_failure; | |
3221 | } | |
543cc10b LE |
3222 | nla_nest_end(skb, nla); |
3223 | return 0; | |
b411b363 | 3224 | |
543cc10b LE |
3225 | nla_put_failure: |
3226 | if (nla) | |
3227 | nla_nest_cancel(skb, nla); | |
3228 | return -EMSGSIZE; | |
3229 | } | |
b411b363 | 3230 | |
a55bbd37 AG |
3231 | /* |
3232 | * The generic netlink dump callbacks are called outside the genl_lock(), so | |
3233 | * they cannot use the simple attribute parsing code which uses global | |
3234 | * attribute tables. | |
3235 | */ | |
3236 | static struct nlattr *find_cfg_context_attr(const struct nlmsghdr *nlh, int attr) | |
3237 | { | |
3238 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
3239 | const int maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; | |
3240 | struct nlattr *nla; | |
3241 | ||
3242 | nla = nla_find(nlmsg_attrdata(nlh, hdrlen), nlmsg_attrlen(nlh, hdrlen), | |
3243 | DRBD_NLA_CFG_CONTEXT); | |
3244 | if (!nla) | |
3245 | return NULL; | |
3246 | return drbd_nla_find_nested(maxtype, nla, __nla_type(attr)); | |
3247 | } | |
3248 | ||
3249 | static void resource_to_info(struct resource_info *, struct drbd_resource *); | |
3250 | ||
3251 | int drbd_adm_dump_resources(struct sk_buff *skb, struct netlink_callback *cb) | |
3252 | { | |
3253 | struct drbd_genlmsghdr *dh; | |
3254 | struct drbd_resource *resource; | |
3255 | struct resource_info resource_info; | |
3256 | struct resource_statistics resource_statistics; | |
3257 | int err; | |
3258 | ||
3259 | rcu_read_lock(); | |
3260 | if (cb->args[0]) { | |
3261 | for_each_resource_rcu(resource, &drbd_resources) | |
3262 | if (resource == (struct drbd_resource *)cb->args[0]) | |
3263 | goto found_resource; | |
3264 | err = 0; /* resource was probably deleted */ | |
3265 | goto out; | |
3266 | } | |
3267 | resource = list_entry(&drbd_resources, | |
3268 | struct drbd_resource, resources); | |
3269 | ||
3270 | found_resource: | |
3271 | list_for_each_entry_continue_rcu(resource, &drbd_resources, resources) { | |
3272 | goto put_result; | |
3273 | } | |
3274 | err = 0; | |
3275 | goto out; | |
3276 | ||
3277 | put_result: | |
3278 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3279 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3280 | NLM_F_MULTI, DRBD_ADM_GET_RESOURCES); | |
3281 | err = -ENOMEM; | |
3282 | if (!dh) | |
3283 | goto out; | |
3284 | dh->minor = -1U; | |
3285 | dh->ret_code = NO_ERROR; | |
3286 | err = nla_put_drbd_cfg_context(skb, resource, NULL, NULL); | |
3287 | if (err) | |
3288 | goto out; | |
3289 | err = res_opts_to_skb(skb, &resource->res_opts, !capable(CAP_SYS_ADMIN)); | |
3290 | if (err) | |
3291 | goto out; | |
3292 | resource_to_info(&resource_info, resource); | |
3293 | err = resource_info_to_skb(skb, &resource_info, !capable(CAP_SYS_ADMIN)); | |
3294 | if (err) | |
3295 | goto out; | |
3296 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
3297 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
3298 | if (err) | |
3299 | goto out; | |
3300 | cb->args[0] = (long)resource; | |
3301 | genlmsg_end(skb, dh); | |
3302 | err = 0; | |
3303 | ||
3304 | out: | |
3305 | rcu_read_unlock(); | |
3306 | if (err) | |
3307 | return err; | |
3308 | return skb->len; | |
3309 | } | |
3310 | ||
3311 | static void device_to_statistics(struct device_statistics *s, | |
3312 | struct drbd_device *device) | |
3313 | { | |
3314 | memset(s, 0, sizeof(*s)); | |
3315 | s->dev_upper_blocked = !may_inc_ap_bio(device); | |
3316 | if (get_ldev(device)) { | |
3317 | struct drbd_md *md = &device->ldev->md; | |
3318 | u64 *history_uuids = (u64 *)s->history_uuids; | |
3319 | struct request_queue *q; | |
3320 | int n; | |
3321 | ||
3322 | spin_lock_irq(&md->uuid_lock); | |
3323 | s->dev_current_uuid = md->uuid[UI_CURRENT]; | |
3324 | BUILD_BUG_ON(sizeof(s->history_uuids) < UI_HISTORY_END - UI_HISTORY_START + 1); | |
3325 | for (n = 0; n < UI_HISTORY_END - UI_HISTORY_START + 1; n++) | |
3326 | history_uuids[n] = md->uuid[UI_HISTORY_START + n]; | |
3327 | for (; n < HISTORY_UUIDS; n++) | |
3328 | history_uuids[n] = 0; | |
3329 | s->history_uuids_len = HISTORY_UUIDS; | |
3330 | spin_unlock_irq(&md->uuid_lock); | |
3331 | ||
3332 | s->dev_disk_flags = md->flags; | |
3333 | q = bdev_get_queue(device->ldev->backing_bdev); | |
3334 | s->dev_lower_blocked = | |
dc3b17cc | 3335 | bdi_congested(q->backing_dev_info, |
a55bbd37 AG |
3336 | (1 << WB_async_congested) | |
3337 | (1 << WB_sync_congested)); | |
3338 | put_ldev(device); | |
3339 | } | |
3340 | s->dev_size = drbd_get_capacity(device->this_bdev); | |
3341 | s->dev_read = device->read_cnt; | |
3342 | s->dev_write = device->writ_cnt; | |
3343 | s->dev_al_writes = device->al_writ_cnt; | |
3344 | s->dev_bm_writes = device->bm_writ_cnt; | |
3345 | s->dev_upper_pending = atomic_read(&device->ap_bio_cnt); | |
3346 | s->dev_lower_pending = atomic_read(&device->local_cnt); | |
3347 | s->dev_al_suspended = test_bit(AL_SUSPENDED, &device->flags); | |
3348 | s->dev_exposed_data_uuid = device->ed_uuid; | |
3349 | } | |
3350 | ||
3351 | static int put_resource_in_arg0(struct netlink_callback *cb, int holder_nr) | |
3352 | { | |
3353 | if (cb->args[0]) { | |
3354 | struct drbd_resource *resource = | |
3355 | (struct drbd_resource *)cb->args[0]; | |
3356 | kref_put(&resource->kref, drbd_destroy_resource); | |
3357 | } | |
3358 | ||
3359 | return 0; | |
3360 | } | |
3361 | ||
3362 | int drbd_adm_dump_devices_done(struct netlink_callback *cb) { | |
3363 | return put_resource_in_arg0(cb, 7); | |
3364 | } | |
3365 | ||
3366 | static void device_to_info(struct device_info *, struct drbd_device *); | |
3367 | ||
3368 | int drbd_adm_dump_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3369 | { | |
3370 | struct nlattr *resource_filter; | |
3371 | struct drbd_resource *resource; | |
3372 | struct drbd_device *uninitialized_var(device); | |
3373 | int minor, err, retcode; | |
3374 | struct drbd_genlmsghdr *dh; | |
3375 | struct device_info device_info; | |
3376 | struct device_statistics device_statistics; | |
3377 | struct idr *idr_to_search; | |
3378 | ||
3379 | resource = (struct drbd_resource *)cb->args[0]; | |
3380 | if (!cb->args[0] && !cb->args[1]) { | |
3381 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3382 | if (resource_filter) { | |
3383 | retcode = ERR_RES_NOT_KNOWN; | |
3384 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3385 | if (!resource) | |
3386 | goto put_result; | |
3387 | cb->args[0] = (long)resource; | |
3388 | } | |
3389 | } | |
3390 | ||
3391 | rcu_read_lock(); | |
3392 | minor = cb->args[1]; | |
3393 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3394 | device = idr_get_next(idr_to_search, &minor); | |
3395 | if (!device) { | |
3396 | err = 0; | |
3397 | goto out; | |
3398 | } | |
3399 | idr_for_each_entry_continue(idr_to_search, device, minor) { | |
3400 | retcode = NO_ERROR; | |
3401 | goto put_result; /* only one iteration */ | |
3402 | } | |
3403 | err = 0; | |
3404 | goto out; /* no more devices */ | |
3405 | ||
3406 | put_result: | |
3407 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3408 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3409 | NLM_F_MULTI, DRBD_ADM_GET_DEVICES); | |
3410 | err = -ENOMEM; | |
3411 | if (!dh) | |
3412 | goto out; | |
3413 | dh->ret_code = retcode; | |
3414 | dh->minor = -1U; | |
3415 | if (retcode == NO_ERROR) { | |
3416 | dh->minor = device->minor; | |
3417 | err = nla_put_drbd_cfg_context(skb, device->resource, NULL, device); | |
3418 | if (err) | |
3419 | goto out; | |
3420 | if (get_ldev(device)) { | |
3421 | struct disk_conf *disk_conf = | |
3422 | rcu_dereference(device->ldev->disk_conf); | |
3423 | ||
3424 | err = disk_conf_to_skb(skb, disk_conf, !capable(CAP_SYS_ADMIN)); | |
3425 | put_ldev(device); | |
3426 | if (err) | |
3427 | goto out; | |
3428 | } | |
3429 | device_to_info(&device_info, device); | |
3430 | err = device_info_to_skb(skb, &device_info, !capable(CAP_SYS_ADMIN)); | |
3431 | if (err) | |
3432 | goto out; | |
3433 | ||
3434 | device_to_statistics(&device_statistics, device); | |
3435 | err = device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
3436 | if (err) | |
3437 | goto out; | |
3438 | cb->args[1] = minor + 1; | |
3439 | } | |
3440 | genlmsg_end(skb, dh); | |
3441 | err = 0; | |
3442 | ||
3443 | out: | |
3444 | rcu_read_unlock(); | |
3445 | if (err) | |
3446 | return err; | |
3447 | return skb->len; | |
3448 | } | |
3449 | ||
3450 | int drbd_adm_dump_connections_done(struct netlink_callback *cb) | |
3451 | { | |
3452 | return put_resource_in_arg0(cb, 6); | |
3453 | } | |
3454 | ||
3455 | enum { SINGLE_RESOURCE, ITERATE_RESOURCES }; | |
3456 | ||
3457 | int drbd_adm_dump_connections(struct sk_buff *skb, struct netlink_callback *cb) | |
3458 | { | |
3459 | struct nlattr *resource_filter; | |
3460 | struct drbd_resource *resource = NULL, *next_resource; | |
3461 | struct drbd_connection *uninitialized_var(connection); | |
3462 | int err = 0, retcode; | |
3463 | struct drbd_genlmsghdr *dh; | |
3464 | struct connection_info connection_info; | |
3465 | struct connection_statistics connection_statistics; | |
3466 | ||
3467 | rcu_read_lock(); | |
3468 | resource = (struct drbd_resource *)cb->args[0]; | |
3469 | if (!cb->args[0]) { | |
3470 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3471 | if (resource_filter) { | |
3472 | retcode = ERR_RES_NOT_KNOWN; | |
3473 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3474 | if (!resource) | |
3475 | goto put_result; | |
3476 | cb->args[0] = (long)resource; | |
3477 | cb->args[1] = SINGLE_RESOURCE; | |
3478 | } | |
3479 | } | |
3480 | if (!resource) { | |
3481 | if (list_empty(&drbd_resources)) | |
3482 | goto out; | |
3483 | resource = list_first_entry(&drbd_resources, struct drbd_resource, resources); | |
3484 | kref_get(&resource->kref); | |
3485 | cb->args[0] = (long)resource; | |
3486 | cb->args[1] = ITERATE_RESOURCES; | |
3487 | } | |
3488 | ||
3489 | next_resource: | |
3490 | rcu_read_unlock(); | |
3491 | mutex_lock(&resource->conf_update); | |
3492 | rcu_read_lock(); | |
3493 | if (cb->args[2]) { | |
3494 | for_each_connection_rcu(connection, resource) | |
3495 | if (connection == (struct drbd_connection *)cb->args[2]) | |
3496 | goto found_connection; | |
3497 | /* connection was probably deleted */ | |
3498 | goto no_more_connections; | |
3499 | } | |
3500 | connection = list_entry(&resource->connections, struct drbd_connection, connections); | |
3501 | ||
3502 | found_connection: | |
3503 | list_for_each_entry_continue_rcu(connection, &resource->connections, connections) { | |
3504 | if (!has_net_conf(connection)) | |
3505 | continue; | |
3506 | retcode = NO_ERROR; | |
3507 | goto put_result; /* only one iteration */ | |
3508 | } | |
3509 | ||
3510 | no_more_connections: | |
3511 | if (cb->args[1] == ITERATE_RESOURCES) { | |
3512 | for_each_resource_rcu(next_resource, &drbd_resources) { | |
3513 | if (next_resource == resource) | |
3514 | goto found_resource; | |
3515 | } | |
3516 | /* resource was probably deleted */ | |
3517 | } | |
3518 | goto out; | |
3519 | ||
3520 | found_resource: | |
3521 | list_for_each_entry_continue_rcu(next_resource, &drbd_resources, resources) { | |
3522 | mutex_unlock(&resource->conf_update); | |
3523 | kref_put(&resource->kref, drbd_destroy_resource); | |
3524 | resource = next_resource; | |
3525 | kref_get(&resource->kref); | |
3526 | cb->args[0] = (long)resource; | |
3527 | cb->args[2] = 0; | |
3528 | goto next_resource; | |
3529 | } | |
3530 | goto out; /* no more resources */ | |
3531 | ||
3532 | put_result: | |
3533 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3534 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3535 | NLM_F_MULTI, DRBD_ADM_GET_CONNECTIONS); | |
3536 | err = -ENOMEM; | |
3537 | if (!dh) | |
3538 | goto out; | |
3539 | dh->ret_code = retcode; | |
3540 | dh->minor = -1U; | |
3541 | if (retcode == NO_ERROR) { | |
3542 | struct net_conf *net_conf; | |
3543 | ||
3544 | err = nla_put_drbd_cfg_context(skb, resource, connection, NULL); | |
3545 | if (err) | |
3546 | goto out; | |
3547 | net_conf = rcu_dereference(connection->net_conf); | |
3548 | if (net_conf) { | |
3549 | err = net_conf_to_skb(skb, net_conf, !capable(CAP_SYS_ADMIN)); | |
3550 | if (err) | |
3551 | goto out; | |
3552 | } | |
3553 | connection_to_info(&connection_info, connection); | |
3554 | err = connection_info_to_skb(skb, &connection_info, !capable(CAP_SYS_ADMIN)); | |
3555 | if (err) | |
3556 | goto out; | |
3557 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
3558 | err = connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
3559 | if (err) | |
3560 | goto out; | |
3561 | cb->args[2] = (long)connection; | |
3562 | } | |
3563 | genlmsg_end(skb, dh); | |
3564 | err = 0; | |
3565 | ||
3566 | out: | |
3567 | rcu_read_unlock(); | |
3568 | if (resource) | |
3569 | mutex_unlock(&resource->conf_update); | |
3570 | if (err) | |
3571 | return err; | |
3572 | return skb->len; | |
3573 | } | |
3574 | ||
3575 | enum mdf_peer_flag { | |
3576 | MDF_PEER_CONNECTED = 1 << 0, | |
3577 | MDF_PEER_OUTDATED = 1 << 1, | |
3578 | MDF_PEER_FENCING = 1 << 2, | |
3579 | MDF_PEER_FULL_SYNC = 1 << 3, | |
3580 | }; | |
3581 | ||
3582 | static void peer_device_to_statistics(struct peer_device_statistics *s, | |
3583 | struct drbd_peer_device *peer_device) | |
3584 | { | |
3585 | struct drbd_device *device = peer_device->device; | |
3586 | ||
3587 | memset(s, 0, sizeof(*s)); | |
3588 | s->peer_dev_received = device->recv_cnt; | |
3589 | s->peer_dev_sent = device->send_cnt; | |
3590 | s->peer_dev_pending = atomic_read(&device->ap_pending_cnt) + | |
3591 | atomic_read(&device->rs_pending_cnt); | |
3592 | s->peer_dev_unacked = atomic_read(&device->unacked_cnt); | |
3593 | s->peer_dev_out_of_sync = drbd_bm_total_weight(device) << (BM_BLOCK_SHIFT - 9); | |
3594 | s->peer_dev_resync_failed = device->rs_failed << (BM_BLOCK_SHIFT - 9); | |
3595 | if (get_ldev(device)) { | |
3596 | struct drbd_md *md = &device->ldev->md; | |
3597 | ||
3598 | spin_lock_irq(&md->uuid_lock); | |
3599 | s->peer_dev_bitmap_uuid = md->uuid[UI_BITMAP]; | |
3600 | spin_unlock_irq(&md->uuid_lock); | |
3601 | s->peer_dev_flags = | |
3602 | (drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND) ? | |
3603 | MDF_PEER_CONNECTED : 0) + | |
3604 | (drbd_md_test_flag(device->ldev, MDF_CONSISTENT) && | |
3605 | !drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE) ? | |
3606 | MDF_PEER_OUTDATED : 0) + | |
3607 | /* FIXME: MDF_PEER_FENCING? */ | |
3608 | (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) ? | |
3609 | MDF_PEER_FULL_SYNC : 0); | |
3610 | put_ldev(device); | |
3611 | } | |
3612 | } | |
3613 | ||
3614 | int drbd_adm_dump_peer_devices_done(struct netlink_callback *cb) | |
3615 | { | |
3616 | return put_resource_in_arg0(cb, 9); | |
3617 | } | |
3618 | ||
3619 | int drbd_adm_dump_peer_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3620 | { | |
3621 | struct nlattr *resource_filter; | |
3622 | struct drbd_resource *resource; | |
3623 | struct drbd_device *uninitialized_var(device); | |
3624 | struct drbd_peer_device *peer_device = NULL; | |
3625 | int minor, err, retcode; | |
3626 | struct drbd_genlmsghdr *dh; | |
3627 | struct idr *idr_to_search; | |
3628 | ||
3629 | resource = (struct drbd_resource *)cb->args[0]; | |
3630 | if (!cb->args[0] && !cb->args[1]) { | |
3631 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3632 | if (resource_filter) { | |
3633 | retcode = ERR_RES_NOT_KNOWN; | |
3634 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3635 | if (!resource) | |
3636 | goto put_result; | |
3637 | } | |
3638 | cb->args[0] = (long)resource; | |
3639 | } | |
3640 | ||
3641 | rcu_read_lock(); | |
3642 | minor = cb->args[1]; | |
3643 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3644 | device = idr_find(idr_to_search, minor); | |
3645 | if (!device) { | |
3646 | next_device: | |
3647 | minor++; | |
3648 | cb->args[2] = 0; | |
3649 | device = idr_get_next(idr_to_search, &minor); | |
3650 | if (!device) { | |
3651 | err = 0; | |
3652 | goto out; | |
3653 | } | |
3654 | } | |
3655 | if (cb->args[2]) { | |
3656 | for_each_peer_device(peer_device, device) | |
3657 | if (peer_device == (struct drbd_peer_device *)cb->args[2]) | |
3658 | goto found_peer_device; | |
3659 | /* peer device was probably deleted */ | |
3660 | goto next_device; | |
3661 | } | |
3662 | /* Make peer_device point to the list head (not the first entry). */ | |
3663 | peer_device = list_entry(&device->peer_devices, struct drbd_peer_device, peer_devices); | |
3664 | ||
3665 | found_peer_device: | |
3666 | list_for_each_entry_continue_rcu(peer_device, &device->peer_devices, peer_devices) { | |
3667 | if (!has_net_conf(peer_device->connection)) | |
3668 | continue; | |
3669 | retcode = NO_ERROR; | |
3670 | goto put_result; /* only one iteration */ | |
3671 | } | |
3672 | goto next_device; | |
3673 | ||
3674 | put_result: | |
3675 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3676 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3677 | NLM_F_MULTI, DRBD_ADM_GET_PEER_DEVICES); | |
3678 | err = -ENOMEM; | |
3679 | if (!dh) | |
3680 | goto out; | |
3681 | dh->ret_code = retcode; | |
3682 | dh->minor = -1U; | |
3683 | if (retcode == NO_ERROR) { | |
3684 | struct peer_device_info peer_device_info; | |
3685 | struct peer_device_statistics peer_device_statistics; | |
3686 | ||
3687 | dh->minor = minor; | |
3688 | err = nla_put_drbd_cfg_context(skb, device->resource, peer_device->connection, device); | |
3689 | if (err) | |
3690 | goto out; | |
3691 | peer_device_to_info(&peer_device_info, peer_device); | |
3692 | err = peer_device_info_to_skb(skb, &peer_device_info, !capable(CAP_SYS_ADMIN)); | |
3693 | if (err) | |
3694 | goto out; | |
3695 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
3696 | err = peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
3697 | if (err) | |
3698 | goto out; | |
3699 | cb->args[1] = minor; | |
3700 | cb->args[2] = (long)peer_device; | |
3701 | } | |
3702 | genlmsg_end(skb, dh); | |
3703 | err = 0; | |
3704 | ||
3705 | out: | |
3706 | rcu_read_unlock(); | |
3707 | if (err) | |
3708 | return err; | |
3709 | return skb->len; | |
3710 | } | |
251b8f8e AG |
3711 | /* |
3712 | * Return the connection of @resource if @resource has exactly one connection. | |
3713 | */ | |
3714 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
3715 | { | |
3716 | struct list_head *connections = &resource->connections; | |
3717 | ||
3718 | if (list_empty(connections) || connections->next->next != connections) | |
3719 | return NULL; | |
3720 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
3721 | } | |
3722 | ||
8ce953aa | 3723 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 3724 | const struct sib_info *sib) |
b411b363 | 3725 | { |
251b8f8e | 3726 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
3727 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
3728 | struct nlattr *nla; | |
3729 | int got_ldev; | |
3b98c0c2 LE |
3730 | int err = 0; |
3731 | int exclude_sensitive; | |
3732 | ||
3733 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
3734 | * to. So we better exclude_sensitive information. | |
3735 | * | |
3736 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
3737 | * in the context of the requesting user process. Exclude sensitive | |
3738 | * information, unless current has superuser. | |
3739 | * | |
3740 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
3741 | * relies on the current implementation of netlink_dump(), which | |
3742 | * executes the dump callback successively from netlink_recvmsg(), | |
3743 | * always in the context of the receiving process */ | |
3744 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
3745 | ||
b30ab791 | 3746 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
3747 | |
3748 | /* We need to add connection name and volume number information still. | |
3749 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 3750 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 3751 | goto nla_put_failure; |
3b98c0c2 | 3752 | |
eb6bea67 | 3753 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
3754 | goto nla_put_failure; |
3755 | ||
daeda1cc | 3756 | rcu_read_lock(); |
f9eb7bf4 AG |
3757 | if (got_ldev) { |
3758 | struct disk_conf *disk_conf; | |
44ed167d | 3759 | |
b30ab791 | 3760 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
3761 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
3762 | } | |
3763 | if (!err) { | |
3764 | struct net_conf *nc; | |
3765 | ||
a6b32bc3 | 3766 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
3767 | if (nc) |
3768 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
3769 | } | |
44ed167d PR |
3770 | rcu_read_unlock(); |
3771 | if (err) | |
3772 | goto nla_put_failure; | |
3b98c0c2 | 3773 | |
3b98c0c2 LE |
3774 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
3775 | if (!nla) | |
3776 | goto nla_put_failure; | |
26ec9287 | 3777 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 | 3778 | nla_put_u32(skb, T_current_state, device->state.i) || |
1dee3f59 ND |
3779 | nla_put_u64_0pad(skb, T_ed_uuid, device->ed_uuid) || |
3780 | nla_put_u64_0pad(skb, T_capacity, | |
3781 | drbd_get_capacity(device->this_bdev)) || | |
3782 | nla_put_u64_0pad(skb, T_send_cnt, device->send_cnt) || | |
3783 | nla_put_u64_0pad(skb, T_recv_cnt, device->recv_cnt) || | |
3784 | nla_put_u64_0pad(skb, T_read_cnt, device->read_cnt) || | |
3785 | nla_put_u64_0pad(skb, T_writ_cnt, device->writ_cnt) || | |
3786 | nla_put_u64_0pad(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
3787 | nla_put_u64_0pad(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
b30ab791 AG |
3788 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || |
3789 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
3790 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 3791 | goto nla_put_failure; |
3b98c0c2 LE |
3792 | |
3793 | if (got_ldev) { | |
39a1aa7f | 3794 | int err; |
b411b363 | 3795 | |
b30ab791 AG |
3796 | spin_lock_irq(&device->ldev->md.uuid_lock); |
3797 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
3798 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
3799 | |
3800 | if (err) | |
3801 | goto nla_put_failure; | |
3802 | ||
b30ab791 | 3803 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
1dee3f59 ND |
3804 | nla_put_u64_0pad(skb, T_bits_total, drbd_bm_bits(device)) || |
3805 | nla_put_u64_0pad(skb, T_bits_oos, | |
3806 | drbd_bm_total_weight(device))) | |
26ec9287 | 3807 | goto nla_put_failure; |
b30ab791 AG |
3808 | if (C_SYNC_SOURCE <= device->state.conn && |
3809 | C_PAUSED_SYNC_T >= device->state.conn) { | |
1dee3f59 ND |
3810 | if (nla_put_u64_0pad(skb, T_bits_rs_total, |
3811 | device->rs_total) || | |
3812 | nla_put_u64_0pad(skb, T_bits_rs_failed, | |
3813 | device->rs_failed)) | |
26ec9287 | 3814 | goto nla_put_failure; |
3b98c0c2 | 3815 | } |
b411b363 | 3816 | } |
b411b363 | 3817 | |
3b98c0c2 LE |
3818 | if (sib) { |
3819 | switch(sib->sib_reason) { | |
3820 | case SIB_SYNC_PROGRESS: | |
3821 | case SIB_GET_STATUS_REPLY: | |
3822 | break; | |
3823 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3824 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3825 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3826 | goto nla_put_failure; | |
3b98c0c2 LE |
3827 | break; |
3828 | case SIB_HELPER_POST: | |
26ec9287 AG |
3829 | if (nla_put_u32(skb, T_helper_exit_code, |
3830 | sib->helper_exit_code)) | |
3831 | goto nla_put_failure; | |
3b98c0c2 LE |
3832 | /* fall through */ |
3833 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3834 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3835 | goto nla_put_failure; | |
3b98c0c2 LE |
3836 | break; |
3837 | } | |
b411b363 | 3838 | } |
3b98c0c2 | 3839 | nla_nest_end(skb, nla); |
b411b363 | 3840 | |
3b98c0c2 LE |
3841 | if (0) |
3842 | nla_put_failure: | |
3843 | err = -EMSGSIZE; | |
3844 | if (got_ldev) | |
b30ab791 | 3845 | put_ldev(device); |
3b98c0c2 | 3846 | return err; |
b411b363 PR |
3847 | } |
3848 | ||
3b98c0c2 | 3849 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3850 | { |
a910b123 | 3851 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3852 | enum drbd_ret_code retcode; |
3853 | int err; | |
b411b363 | 3854 | |
a910b123 | 3855 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3856 | if (!adm_ctx.reply_skb) |
3857 | return retcode; | |
3858 | if (retcode != NO_ERROR) | |
3859 | goto out; | |
b411b363 | 3860 | |
b30ab791 | 3861 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3862 | if (err) { |
3863 | nlmsg_free(adm_ctx.reply_skb); | |
3864 | return err; | |
b411b363 | 3865 | } |
3b98c0c2 | 3866 | out: |
a910b123 | 3867 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3868 | return 0; |
b411b363 PR |
3869 | } |
3870 | ||
4b7a530f | 3871 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3872 | { |
b30ab791 | 3873 | struct drbd_device *device; |
3b98c0c2 | 3874 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3875 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3876 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3877 | struct drbd_resource *tmp; |
543cc10b LE |
3878 | unsigned volume = cb->args[1]; |
3879 | ||
3880 | /* Open coded, deferred, iteration: | |
77c556f6 | 3881 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3882 | * connection = "first connection of resource or undefined"; |
3883 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3884 | * ... |
3885 | * } | |
3886 | * } | |
77c556f6 | 3887 | * where resource is cb->args[0]; |
543cc10b LE |
3888 | * and i is cb->args[1]; |
3889 | * | |
71932efc LE |
3890 | * cb->args[2] indicates if we shall loop over all resources, |
3891 | * or just dump all volumes of a single resource. | |
3892 | * | |
3b98c0c2 LE |
3893 | * This may miss entries inserted after this dump started, |
3894 | * or entries deleted before they are reached. | |
543cc10b | 3895 | * |
b30ab791 | 3896 | * We need to make sure the device won't disappear while |
543cc10b LE |
3897 | * we are looking at it, and revalidate our iterators |
3898 | * on each iteration. | |
3899 | */ | |
b411b363 | 3900 | |
05a10ec7 | 3901 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3902 | rcu_read_lock(); |
543cc10b | 3903 | /* revalidate iterator position */ |
77c556f6 | 3904 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3905 | if (pos == NULL) { |
3906 | /* first iteration */ | |
3907 | pos = tmp; | |
77c556f6 | 3908 | resource = pos; |
543cc10b LE |
3909 | break; |
3910 | } | |
3911 | if (tmp == pos) { | |
77c556f6 | 3912 | resource = pos; |
543cc10b LE |
3913 | break; |
3914 | } | |
b411b363 | 3915 | } |
77c556f6 AG |
3916 | if (resource) { |
3917 | next_resource: | |
251b8f8e AG |
3918 | device = idr_get_next(&resource->devices, &volume); |
3919 | if (!device) { | |
77c556f6 AG |
3920 | /* No more volumes to dump on this resource. |
3921 | * Advance resource iterator. */ | |
3922 | pos = list_entry_rcu(resource->resources.next, | |
3923 | struct drbd_resource, resources); | |
3924 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3925 | if (volume != 0) { |
71932efc LE |
3926 | /* If we reached the end of the list, |
3927 | * or only a single resource dump was requested, | |
3928 | * we are done. */ | |
77c556f6 | 3929 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3930 | goto out; |
543cc10b | 3931 | volume = 0; |
77c556f6 AG |
3932 | resource = pos; |
3933 | goto next_resource; | |
543cc10b LE |
3934 | } |
3935 | } | |
3936 | ||
98683650 | 3937 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3938 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3939 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3940 | if (!dh) | |
543cc10b LE |
3941 | goto out; |
3942 | ||
251b8f8e | 3943 | if (!device) { |
bde89a9e | 3944 | /* This is a connection without a single volume. |
367d675d LE |
3945 | * Suprisingly enough, it may have a network |
3946 | * configuration. */ | |
251b8f8e AG |
3947 | struct drbd_connection *connection; |
3948 | ||
543cc10b LE |
3949 | dh->minor = -1U; |
3950 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3951 | connection = the_only_connection(resource); |
3952 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3953 | goto cancel; |
251b8f8e AG |
3954 | if (connection) { |
3955 | struct net_conf *nc; | |
3956 | ||
3957 | nc = rcu_dereference(connection->net_conf); | |
3958 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3959 | goto cancel; | |
3960 | } | |
367d675d | 3961 | goto done; |
543cc10b | 3962 | } |
b411b363 | 3963 | |
0b0ba1ef | 3964 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3965 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3966 | |
b30ab791 | 3967 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3968 | dh->ret_code = NO_ERROR; |
3969 | ||
b30ab791 | 3970 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3971 | cancel: |
3b98c0c2 | 3972 | genlmsg_cancel(skb, dh); |
543cc10b | 3973 | goto out; |
3b98c0c2 | 3974 | } |
367d675d | 3975 | done: |
3b98c0c2 | 3976 | genlmsg_end(skb, dh); |
bde89a9e | 3977 | } |
b411b363 | 3978 | |
543cc10b | 3979 | out: |
c141ebda | 3980 | rcu_read_unlock(); |
543cc10b | 3981 | /* where to start the next iteration */ |
bde89a9e | 3982 | cb->args[0] = (long)pos; |
77c556f6 | 3983 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3984 | |
77c556f6 | 3985 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3986 | * Which will terminate the dump. */ |
3b98c0c2 | 3987 | return skb->len; |
b411b363 PR |
3988 | } |
3989 | ||
71932efc LE |
3990 | /* |
3991 | * Request status of all resources, or of all volumes within a single resource. | |
3992 | * | |
3993 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3994 | * Which means we cannot use the family->attrbuf or other such members, because | |
3995 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3996 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3997 | * | |
3998 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3999 | */ |
71932efc | 4000 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 4001 | { |
71932efc LE |
4002 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
4003 | struct nlattr *nla; | |
7c3063cc | 4004 | const char *resource_name; |
4bc76048 | 4005 | struct drbd_resource *resource; |
7c3063cc | 4006 | int maxtype; |
71932efc LE |
4007 | |
4008 | /* Is this a followup call? */ | |
4009 | if (cb->args[0]) { | |
4010 | /* ... of a single resource dump, | |
4011 | * and the resource iterator has been advanced already? */ | |
4012 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
4013 | return 0; /* DONE. */ | |
4014 | goto dump; | |
4015 | } | |
4016 | ||
4017 | /* First call (from netlink_dump_start). We need to figure out | |
4018 | * which resource(s) the user wants us to dump. */ | |
4019 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
4020 | nlmsg_attrlen(cb->nlh, hdrlen), | |
4021 | DRBD_NLA_CFG_CONTEXT); | |
4022 | ||
4023 | /* No explicit context given. Dump all. */ | |
4024 | if (!nla) | |
4025 | goto dump; | |
7c3063cc AG |
4026 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
4027 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
4028 | if (IS_ERR(nla)) | |
4029 | return PTR_ERR(nla); | |
71932efc LE |
4030 | /* context given, but no name present? */ |
4031 | if (!nla) | |
4032 | return -EINVAL; | |
7c3063cc | 4033 | resource_name = nla_data(nla); |
4bc76048 AG |
4034 | if (!*resource_name) |
4035 | return -ENODEV; | |
4036 | resource = drbd_find_resource(resource_name); | |
4037 | if (!resource) | |
71932efc LE |
4038 | return -ENODEV; |
4039 | ||
4bc76048 | 4040 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 4041 | |
71932efc | 4042 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 4043 | * only dump this connection. */ |
4bc76048 | 4044 | cb->args[0] = (long)resource; |
71932efc | 4045 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 4046 | cb->args[2] = (long)resource; |
71932efc LE |
4047 | |
4048 | dump: | |
4049 | return get_one_status(skb, cb); | |
4050 | } | |
b411b363 | 4051 | |
3b98c0c2 | 4052 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4053 | { |
a910b123 | 4054 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
4055 | enum drbd_ret_code retcode; |
4056 | struct timeout_parms tp; | |
4057 | int err; | |
b411b363 | 4058 | |
a910b123 | 4059 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4060 | if (!adm_ctx.reply_skb) |
4061 | return retcode; | |
4062 | if (retcode != NO_ERROR) | |
4063 | goto out; | |
b411b363 | 4064 | |
3b98c0c2 | 4065 | tp.timeout_type = |
b30ab791 AG |
4066 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
4067 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 4068 | UT_DEFAULT; |
b411b363 | 4069 | |
3b98c0c2 LE |
4070 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
4071 | if (err) { | |
4072 | nlmsg_free(adm_ctx.reply_skb); | |
4073 | return err; | |
4074 | } | |
4075 | out: | |
a910b123 | 4076 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4077 | return 0; |
b411b363 PR |
4078 | } |
4079 | ||
3b98c0c2 | 4080 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4081 | { |
a910b123 | 4082 | struct drbd_config_context adm_ctx; |
b30ab791 | 4083 | struct drbd_device *device; |
3b98c0c2 | 4084 | enum drbd_ret_code retcode; |
58ffa580 | 4085 | struct start_ov_parms parms; |
b411b363 | 4086 | |
a910b123 | 4087 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4088 | if (!adm_ctx.reply_skb) |
4089 | return retcode; | |
4090 | if (retcode != NO_ERROR) | |
4091 | goto out; | |
873b0d5f | 4092 | |
b30ab791 | 4093 | device = adm_ctx.device; |
58ffa580 LE |
4094 | |
4095 | /* resume from last known position, if possible */ | |
b30ab791 | 4096 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 4097 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 4098 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 4099 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
4100 | if (err) { |
4101 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4102 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
4103 | goto out; |
4104 | } | |
b411b363 | 4105 | } |
9e276872 LE |
4106 | mutex_lock(&adm_ctx.resource->adm_mutex); |
4107 | ||
58ffa580 | 4108 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
4109 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
4110 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
4111 | |
4112 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
4113 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
4114 | drbd_suspend_io(device); |
4115 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
4116 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
4117 | drbd_resume_io(device); | |
9e276872 LE |
4118 | |
4119 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 4120 | out: |
a910b123 | 4121 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4122 | return 0; |
4123 | } | |
4124 | ||
4125 | ||
3b98c0c2 | 4126 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4127 | { |
a910b123 | 4128 | struct drbd_config_context adm_ctx; |
b30ab791 | 4129 | struct drbd_device *device; |
3b98c0c2 | 4130 | enum drbd_ret_code retcode; |
b411b363 PR |
4131 | int skip_initial_sync = 0; |
4132 | int err; | |
3b98c0c2 | 4133 | struct new_c_uuid_parms args; |
b411b363 | 4134 | |
a910b123 | 4135 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4136 | if (!adm_ctx.reply_skb) |
4137 | return retcode; | |
4138 | if (retcode != NO_ERROR) | |
4139 | goto out_nolock; | |
b411b363 | 4140 | |
b30ab791 | 4141 | device = adm_ctx.device; |
3b98c0c2 LE |
4142 | memset(&args, 0, sizeof(args)); |
4143 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 4144 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
4145 | if (err) { |
4146 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4147 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
4148 | goto out_nolock; |
4149 | } | |
b411b363 PR |
4150 | } |
4151 | ||
9e276872 | 4152 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 4153 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 4154 | |
b30ab791 | 4155 | if (!get_ldev(device)) { |
b411b363 PR |
4156 | retcode = ERR_NO_DISK; |
4157 | goto out; | |
4158 | } | |
4159 | ||
4160 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
4161 | if (device->state.conn == C_CONNECTED && |
4162 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 4163 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 4164 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 4165 | skip_initial_sync = 1; |
b30ab791 | 4166 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
4167 | retcode = ERR_CONNECTED; |
4168 | goto out_dec; | |
4169 | } | |
4170 | ||
b30ab791 AG |
4171 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
4172 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
4173 | |
4174 | if (args.clear_bm) { | |
b30ab791 | 4175 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 4176 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 4177 | if (err) { |
d0180171 | 4178 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
4179 | retcode = ERR_IO_MD_DISK; |
4180 | } | |
4181 | if (skip_initial_sync) { | |
69a22773 | 4182 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
4183 | _drbd_uuid_set(device, UI_BITMAP, 0); |
4184 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 4185 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 4186 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 4187 | CS_VERBOSE, NULL); |
0500813f | 4188 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
4189 | } |
4190 | } | |
4191 | ||
b30ab791 | 4192 | drbd_md_sync(device); |
b411b363 | 4193 | out_dec: |
b30ab791 | 4194 | put_ldev(device); |
b411b363 | 4195 | out: |
b30ab791 | 4196 | mutex_unlock(device->state_mutex); |
9e276872 | 4197 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 4198 | out_nolock: |
a910b123 | 4199 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4200 | return 0; |
4201 | } | |
4202 | ||
3b98c0c2 | 4203 | static enum drbd_ret_code |
a910b123 | 4204 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 4205 | { |
a910b123 | 4206 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 4207 | if (!name || !name[0]) { |
a910b123 | 4208 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 4209 | return ERR_MANDATORY_TAG; |
b411b363 | 4210 | } |
3b98c0c2 LE |
4211 | /* if we want to use these in sysfs/configfs/debugfs some day, |
4212 | * we must not allow slashes */ | |
4213 | if (strchr(name, '/')) { | |
a910b123 | 4214 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 4215 | return ERR_INVALID_REQUEST; |
b411b363 | 4216 | } |
3b98c0c2 | 4217 | return NO_ERROR; |
774b3055 | 4218 | } |
b411b363 | 4219 | |
a2972846 AG |
4220 | static void resource_to_info(struct resource_info *info, |
4221 | struct drbd_resource *resource) | |
4222 | { | |
4223 | info->res_role = conn_highest_role(first_connection(resource)); | |
4224 | info->res_susp = resource->susp; | |
4225 | info->res_susp_nod = resource->susp_nod; | |
4226 | info->res_susp_fen = resource->susp_fen; | |
4227 | } | |
4228 | ||
789c1b62 | 4229 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4230 | { |
a2972846 | 4231 | struct drbd_connection *connection; |
a910b123 | 4232 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 4233 | enum drbd_ret_code retcode; |
afbbfa88 AG |
4234 | struct res_opts res_opts; |
4235 | int err; | |
b411b363 | 4236 | |
a910b123 | 4237 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
4238 | if (!adm_ctx.reply_skb) |
4239 | return retcode; | |
4240 | if (retcode != NO_ERROR) | |
4241 | goto out; | |
b411b363 | 4242 | |
afbbfa88 AG |
4243 | set_res_opts_defaults(&res_opts); |
4244 | err = res_opts_from_attrs(&res_opts, info); | |
4245 | if (err && err != -ENOMSG) { | |
4246 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4247 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 4248 | goto out; |
b411b363 PR |
4249 | } |
4250 | ||
a910b123 | 4251 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
4252 | if (retcode != NO_ERROR) |
4253 | goto out; | |
b411b363 | 4254 | |
5c661042 | 4255 | if (adm_ctx.resource) { |
38f19616 LE |
4256 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
4257 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 4258 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
4259 | } |
4260 | /* else: still NO_ERROR */ | |
3b98c0c2 | 4261 | goto out; |
b411b363 | 4262 | } |
b411b363 | 4263 | |
9e276872 | 4264 | /* not yet safe for genl_family.parallel_ops */ |
28bc3b8c | 4265 | mutex_lock(&resources_mutex); |
a2972846 | 4266 | connection = conn_create(adm_ctx.resource_name, &res_opts); |
28bc3b8c | 4267 | mutex_unlock(&resources_mutex); |
a2972846 AG |
4268 | |
4269 | if (connection) { | |
4270 | struct resource_info resource_info; | |
4271 | ||
4272 | mutex_lock(¬ification_mutex); | |
4273 | resource_to_info(&resource_info, connection->resource); | |
4274 | notify_resource_state(NULL, 0, connection->resource, | |
4275 | &resource_info, NOTIFY_CREATE); | |
4276 | mutex_unlock(¬ification_mutex); | |
4277 | } else | |
4278 | retcode = ERR_NOMEM; | |
4279 | ||
3b98c0c2 | 4280 | out: |
a910b123 | 4281 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4282 | return 0; |
b411b363 PR |
4283 | } |
4284 | ||
a2972846 AG |
4285 | static void device_to_info(struct device_info *info, |
4286 | struct drbd_device *device) | |
4287 | { | |
4288 | info->dev_disk_state = device->state.disk; | |
4289 | } | |
4290 | ||
4291 | ||
05a10ec7 | 4292 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4293 | { |
a910b123 | 4294 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
4295 | struct drbd_genlmsghdr *dh = info->userhdr; |
4296 | enum drbd_ret_code retcode; | |
b411b363 | 4297 | |
a910b123 | 4298 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4299 | if (!adm_ctx.reply_skb) |
4300 | return retcode; | |
4301 | if (retcode != NO_ERROR) | |
4302 | goto out; | |
b411b363 | 4303 | |
f2257a56 | 4304 | if (dh->minor > MINORMASK) { |
a910b123 | 4305 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
4306 | retcode = ERR_INVALID_REQUEST; |
4307 | goto out; | |
b411b363 | 4308 | } |
0c8e36d9 | 4309 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 4310 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
4311 | retcode = ERR_INVALID_REQUEST; |
4312 | goto out; | |
b411b363 | 4313 | } |
b411b363 | 4314 | |
38f19616 | 4315 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 4316 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 4317 | if (adm_ctx.device) { |
38f19616 | 4318 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
179e20b8 | 4319 | retcode = ERR_MINOR_OR_VOLUME_EXISTS; |
38f19616 LE |
4320 | /* else: still NO_ERROR */ |
4321 | goto out; | |
b411b363 | 4322 | } |
38f19616 | 4323 | |
9e276872 | 4324 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 4325 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
a2972846 AG |
4326 | if (retcode == NO_ERROR) { |
4327 | struct drbd_device *device; | |
4328 | struct drbd_peer_device *peer_device; | |
4329 | struct device_info info; | |
4330 | unsigned int peer_devices = 0; | |
4331 | enum drbd_notification_type flags; | |
4332 | ||
4333 | device = minor_to_device(dh->minor); | |
4334 | for_each_peer_device(peer_device, device) { | |
4335 | if (!has_net_conf(peer_device->connection)) | |
4336 | continue; | |
4337 | peer_devices++; | |
4338 | } | |
4339 | ||
4340 | device_to_info(&info, device); | |
4341 | mutex_lock(¬ification_mutex); | |
4342 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4343 | notify_device_state(NULL, 0, device, &info, NOTIFY_CREATE | flags); | |
4344 | for_each_peer_device(peer_device, device) { | |
4345 | struct peer_device_info peer_device_info; | |
4346 | ||
4347 | if (!has_net_conf(peer_device->connection)) | |
4348 | continue; | |
4349 | peer_device_to_info(&peer_device_info, peer_device); | |
4350 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4351 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, | |
4352 | NOTIFY_CREATE | flags); | |
4353 | } | |
4354 | mutex_unlock(¬ification_mutex); | |
4355 | } | |
9e276872 | 4356 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 4357 | out: |
a910b123 | 4358 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4359 | return 0; |
b411b363 PR |
4360 | } |
4361 | ||
05a10ec7 | 4362 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 4363 | { |
a2972846 AG |
4364 | struct drbd_peer_device *peer_device; |
4365 | ||
b30ab791 AG |
4366 | if (device->state.disk == D_DISKLESS && |
4367 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
4368 | * we may want to delete a minor from a live replication group. |
4369 | */ | |
b30ab791 | 4370 | device->state.role == R_SECONDARY) { |
a2972846 AG |
4371 | struct drbd_connection *connection = |
4372 | first_connection(device->resource); | |
4373 | ||
b30ab791 | 4374 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), |
369bea63 | 4375 | CS_VERBOSE + CS_WAIT_COMPLETE); |
a2972846 AG |
4376 | |
4377 | /* If the state engine hasn't stopped the sender thread yet, we | |
4378 | * need to flush the sender work queue before generating the | |
4379 | * DESTROY events here. */ | |
4380 | if (get_t_state(&connection->worker) == RUNNING) | |
4381 | drbd_flush_workqueue(&connection->sender_work); | |
4382 | ||
4383 | mutex_lock(¬ification_mutex); | |
4384 | for_each_peer_device(peer_device, device) { | |
4385 | if (!has_net_conf(peer_device->connection)) | |
4386 | continue; | |
4387 | notify_peer_device_state(NULL, 0, peer_device, NULL, | |
4388 | NOTIFY_DESTROY | NOTIFY_CONTINUES); | |
4389 | } | |
4390 | notify_device_state(NULL, 0, device, NULL, NOTIFY_DESTROY); | |
4391 | mutex_unlock(¬ification_mutex); | |
4392 | ||
f82795d6 | 4393 | drbd_delete_device(device); |
85f75dd7 LE |
4394 | return NO_ERROR; |
4395 | } else | |
4396 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
4397 | } |
4398 | ||
05a10ec7 | 4399 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4400 | { |
a910b123 | 4401 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 4402 | enum drbd_ret_code retcode; |
b411b363 | 4403 | |
a910b123 | 4404 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4405 | if (!adm_ctx.reply_skb) |
4406 | return retcode; | |
4407 | if (retcode != NO_ERROR) | |
4408 | goto out; | |
b411b363 | 4409 | |
9e276872 | 4410 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 4411 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 4412 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 4413 | out: |
a910b123 | 4414 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 4415 | return 0; |
b411b363 PR |
4416 | } |
4417 | ||
179e20b8 AG |
4418 | static int adm_del_resource(struct drbd_resource *resource) |
4419 | { | |
4420 | struct drbd_connection *connection; | |
4421 | ||
4422 | for_each_connection(connection, resource) { | |
4423 | if (connection->cstate > C_STANDALONE) | |
4424 | return ERR_NET_CONFIGURED; | |
4425 | } | |
4426 | if (!idr_is_empty(&resource->devices)) | |
4427 | return ERR_RES_IN_USE; | |
4428 | ||
a2972846 AG |
4429 | /* The state engine has stopped the sender thread, so we don't |
4430 | * need to flush the sender work queue before generating the | |
4431 | * DESTROY event here. */ | |
4432 | mutex_lock(¬ification_mutex); | |
4433 | notify_resource_state(NULL, 0, resource, NULL, NOTIFY_DESTROY); | |
4434 | mutex_unlock(¬ification_mutex); | |
4435 | ||
28bc3b8c | 4436 | mutex_lock(&resources_mutex); |
179e20b8 | 4437 | list_del_rcu(&resource->resources); |
28bc3b8c | 4438 | mutex_unlock(&resources_mutex); |
179e20b8 AG |
4439 | /* Make sure all threads have actually stopped: state handling only |
4440 | * does drbd_thread_stop_nowait(). */ | |
4441 | list_for_each_entry(connection, &resource->connections, connections) | |
4442 | drbd_thread_stop(&connection->worker); | |
4443 | synchronize_rcu(); | |
4444 | drbd_free_resource(resource); | |
4445 | return NO_ERROR; | |
4446 | } | |
4447 | ||
85f75dd7 | 4448 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4449 | { |
a910b123 | 4450 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
4451 | struct drbd_resource *resource; |
4452 | struct drbd_connection *connection; | |
4453 | struct drbd_device *device; | |
f3dfa40a | 4454 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 4455 | unsigned i; |
b411b363 | 4456 | |
a910b123 | 4457 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
4458 | if (!adm_ctx.reply_skb) |
4459 | return retcode; | |
4460 | if (retcode != NO_ERROR) | |
9e276872 | 4461 | goto finish; |
b411b363 | 4462 | |
b6f85ef9 | 4463 | resource = adm_ctx.resource; |
9e276872 | 4464 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 4465 | /* demote */ |
b6f85ef9 AG |
4466 | for_each_connection(connection, resource) { |
4467 | struct drbd_peer_device *peer_device; | |
4468 | ||
4469 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
4470 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
4471 | if (retcode < SS_SUCCESS) { | |
a910b123 | 4472 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
4473 | goto out; |
4474 | } | |
4475 | } | |
4476 | ||
4477 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 4478 | if (retcode < SS_SUCCESS) { |
a910b123 | 4479 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 4480 | goto out; |
85f75dd7 | 4481 | } |
b411b363 | 4482 | } |
b411b363 | 4483 | |
85f75dd7 | 4484 | /* detach */ |
b6f85ef9 AG |
4485 | idr_for_each_entry(&resource->devices, device, i) { |
4486 | retcode = adm_detach(device, 0); | |
27012382 | 4487 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 4488 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 4489 | goto out; |
85f75dd7 LE |
4490 | } |
4491 | } | |
b411b363 | 4492 | |
85f75dd7 | 4493 | /* delete volumes */ |
b6f85ef9 AG |
4494 | idr_for_each_entry(&resource->devices, device, i) { |
4495 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
4496 | if (retcode != NO_ERROR) { |
4497 | /* "can not happen" */ | |
a910b123 | 4498 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 4499 | goto out; |
85f75dd7 LE |
4500 | } |
4501 | } | |
b411b363 | 4502 | |
179e20b8 | 4503 | retcode = adm_del_resource(resource); |
3b98c0c2 | 4504 | out: |
9e276872 LE |
4505 | mutex_unlock(&resource->adm_mutex); |
4506 | finish: | |
a910b123 | 4507 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4508 | return 0; |
b411b363 PR |
4509 | } |
4510 | ||
789c1b62 | 4511 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4512 | { |
a910b123 | 4513 | struct drbd_config_context adm_ctx; |
77c556f6 | 4514 | struct drbd_resource *resource; |
3b98c0c2 | 4515 | enum drbd_ret_code retcode; |
b411b363 | 4516 | |
a910b123 | 4517 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4518 | if (!adm_ctx.reply_skb) |
4519 | return retcode; | |
4520 | if (retcode != NO_ERROR) | |
9e276872 | 4521 | goto finish; |
77c556f6 | 4522 | resource = adm_ctx.resource; |
b411b363 | 4523 | |
179e20b8 AG |
4524 | mutex_lock(&resource->adm_mutex); |
4525 | retcode = adm_del_resource(resource); | |
9e276872 LE |
4526 | mutex_unlock(&resource->adm_mutex); |
4527 | finish: | |
a910b123 | 4528 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4529 | return 0; |
4530 | } | |
4531 | ||
b30ab791 | 4532 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 4533 | { |
3b98c0c2 LE |
4534 | struct sk_buff *msg; |
4535 | struct drbd_genlmsghdr *d_out; | |
4536 | unsigned seq; | |
4537 | int err = -ENOMEM; | |
4538 | ||
4539 | seq = atomic_inc_return(&drbd_genl_seq); | |
4540 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4541 | if (!msg) | |
4542 | goto failed; | |
4543 | ||
4544 | err = -EMSGSIZE; | |
4545 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
4546 | if (!d_out) /* cannot happen, but anyways. */ | |
4547 | goto nla_put_failure; | |
b30ab791 | 4548 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 4549 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 4550 | |
b30ab791 | 4551 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
4552 | goto nla_put_failure; |
4553 | genlmsg_end(msg, d_out); | |
d38f8612 | 4554 | err = drbd_genl_multicast_events(msg, GFP_NOWAIT); |
3b98c0c2 LE |
4555 | /* msg has been consumed or freed in netlink_broadcast() */ |
4556 | if (err && err != -ESRCH) | |
4557 | goto failed; | |
b411b363 | 4558 | |
3b98c0c2 | 4559 | return; |
b411b363 | 4560 | |
3b98c0c2 LE |
4561 | nla_put_failure: |
4562 | nlmsg_free(msg); | |
4563 | failed: | |
d0180171 | 4564 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
4565 | "Event seq:%u sib_reason:%u\n", |
4566 | err, seq, sib->sib_reason); | |
b411b363 | 4567 | } |
a2972846 | 4568 | |
a2972846 AG |
4569 | static int nla_put_notification_header(struct sk_buff *msg, |
4570 | enum drbd_notification_type type) | |
4571 | { | |
4572 | struct drbd_notification_header nh = { | |
4573 | .nh_type = type, | |
4574 | }; | |
4575 | ||
4576 | return drbd_notification_header_to_skb(msg, &nh, true); | |
4577 | } | |
4578 | ||
4579 | void notify_resource_state(struct sk_buff *skb, | |
4580 | unsigned int seq, | |
4581 | struct drbd_resource *resource, | |
4582 | struct resource_info *resource_info, | |
4583 | enum drbd_notification_type type) | |
4584 | { | |
4585 | struct resource_statistics resource_statistics; | |
4586 | struct drbd_genlmsghdr *dh; | |
4587 | bool multicast = false; | |
4588 | int err; | |
4589 | ||
4590 | if (!skb) { | |
4591 | seq = atomic_inc_return(¬ify_genl_seq); | |
4592 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4593 | err = -ENOMEM; | |
4594 | if (!skb) | |
4595 | goto failed; | |
4596 | multicast = true; | |
4597 | } | |
4598 | ||
4599 | err = -EMSGSIZE; | |
4600 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_RESOURCE_STATE); | |
4601 | if (!dh) | |
4602 | goto nla_put_failure; | |
4603 | dh->minor = -1U; | |
4604 | dh->ret_code = NO_ERROR; | |
4605 | if (nla_put_drbd_cfg_context(skb, resource, NULL, NULL) || | |
4606 | nla_put_notification_header(skb, type) || | |
4607 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4608 | resource_info_to_skb(skb, resource_info, true))) | |
4609 | goto nla_put_failure; | |
4610 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
4611 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
4612 | if (err) | |
4613 | goto nla_put_failure; | |
4614 | genlmsg_end(skb, dh); | |
4615 | if (multicast) { | |
d38f8612 | 4616 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4617 | /* skb has been consumed or freed in netlink_broadcast() */ |
4618 | if (err && err != -ESRCH) | |
4619 | goto failed; | |
4620 | } | |
4621 | return; | |
4622 | ||
4623 | nla_put_failure: | |
4624 | nlmsg_free(skb); | |
4625 | failed: | |
4626 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4627 | err, seq); | |
4628 | } | |
4629 | ||
4630 | void notify_device_state(struct sk_buff *skb, | |
4631 | unsigned int seq, | |
4632 | struct drbd_device *device, | |
4633 | struct device_info *device_info, | |
4634 | enum drbd_notification_type type) | |
4635 | { | |
4636 | struct device_statistics device_statistics; | |
4637 | struct drbd_genlmsghdr *dh; | |
4638 | bool multicast = false; | |
4639 | int err; | |
4640 | ||
4641 | if (!skb) { | |
4642 | seq = atomic_inc_return(¬ify_genl_seq); | |
4643 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4644 | err = -ENOMEM; | |
4645 | if (!skb) | |
4646 | goto failed; | |
4647 | multicast = true; | |
4648 | } | |
4649 | ||
4650 | err = -EMSGSIZE; | |
4651 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_DEVICE_STATE); | |
4652 | if (!dh) | |
4653 | goto nla_put_failure; | |
4654 | dh->minor = device->minor; | |
4655 | dh->ret_code = NO_ERROR; | |
4656 | if (nla_put_drbd_cfg_context(skb, device->resource, NULL, device) || | |
4657 | nla_put_notification_header(skb, type) || | |
4658 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4659 | device_info_to_skb(skb, device_info, true))) | |
4660 | goto nla_put_failure; | |
4661 | device_to_statistics(&device_statistics, device); | |
4662 | device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
4663 | genlmsg_end(skb, dh); | |
4664 | if (multicast) { | |
d38f8612 | 4665 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4666 | /* skb has been consumed or freed in netlink_broadcast() */ |
4667 | if (err && err != -ESRCH) | |
4668 | goto failed; | |
4669 | } | |
4670 | return; | |
4671 | ||
4672 | nla_put_failure: | |
4673 | nlmsg_free(skb); | |
4674 | failed: | |
4675 | drbd_err(device, "Error %d while broadcasting event. Event seq:%u\n", | |
4676 | err, seq); | |
4677 | } | |
4678 | ||
4679 | void notify_connection_state(struct sk_buff *skb, | |
4680 | unsigned int seq, | |
4681 | struct drbd_connection *connection, | |
4682 | struct connection_info *connection_info, | |
4683 | enum drbd_notification_type type) | |
4684 | { | |
4685 | struct connection_statistics connection_statistics; | |
4686 | struct drbd_genlmsghdr *dh; | |
4687 | bool multicast = false; | |
4688 | int err; | |
4689 | ||
4690 | if (!skb) { | |
4691 | seq = atomic_inc_return(¬ify_genl_seq); | |
4692 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4693 | err = -ENOMEM; | |
4694 | if (!skb) | |
4695 | goto failed; | |
4696 | multicast = true; | |
4697 | } | |
4698 | ||
4699 | err = -EMSGSIZE; | |
4700 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_CONNECTION_STATE); | |
4701 | if (!dh) | |
4702 | goto nla_put_failure; | |
4703 | dh->minor = -1U; | |
4704 | dh->ret_code = NO_ERROR; | |
4705 | if (nla_put_drbd_cfg_context(skb, connection->resource, connection, NULL) || | |
4706 | nla_put_notification_header(skb, type) || | |
4707 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4708 | connection_info_to_skb(skb, connection_info, true))) | |
4709 | goto nla_put_failure; | |
4710 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
4711 | connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
4712 | genlmsg_end(skb, dh); | |
4713 | if (multicast) { | |
d38f8612 | 4714 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4715 | /* skb has been consumed or freed in netlink_broadcast() */ |
4716 | if (err && err != -ESRCH) | |
4717 | goto failed; | |
4718 | } | |
4719 | return; | |
4720 | ||
4721 | nla_put_failure: | |
4722 | nlmsg_free(skb); | |
4723 | failed: | |
4724 | drbd_err(connection, "Error %d while broadcasting event. Event seq:%u\n", | |
4725 | err, seq); | |
4726 | } | |
4727 | ||
4728 | void notify_peer_device_state(struct sk_buff *skb, | |
4729 | unsigned int seq, | |
4730 | struct drbd_peer_device *peer_device, | |
4731 | struct peer_device_info *peer_device_info, | |
4732 | enum drbd_notification_type type) | |
4733 | { | |
4734 | struct peer_device_statistics peer_device_statistics; | |
4735 | struct drbd_resource *resource = peer_device->device->resource; | |
4736 | struct drbd_genlmsghdr *dh; | |
4737 | bool multicast = false; | |
4738 | int err; | |
4739 | ||
4740 | if (!skb) { | |
4741 | seq = atomic_inc_return(¬ify_genl_seq); | |
4742 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4743 | err = -ENOMEM; | |
4744 | if (!skb) | |
4745 | goto failed; | |
4746 | multicast = true; | |
4747 | } | |
4748 | ||
4749 | err = -EMSGSIZE; | |
4750 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_PEER_DEVICE_STATE); | |
4751 | if (!dh) | |
4752 | goto nla_put_failure; | |
4753 | dh->minor = -1U; | |
4754 | dh->ret_code = NO_ERROR; | |
4755 | if (nla_put_drbd_cfg_context(skb, resource, peer_device->connection, peer_device->device) || | |
4756 | nla_put_notification_header(skb, type) || | |
4757 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4758 | peer_device_info_to_skb(skb, peer_device_info, true))) | |
4759 | goto nla_put_failure; | |
4760 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
4761 | peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
4762 | genlmsg_end(skb, dh); | |
4763 | if (multicast) { | |
d38f8612 | 4764 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4765 | /* skb has been consumed or freed in netlink_broadcast() */ |
4766 | if (err && err != -ESRCH) | |
4767 | goto failed; | |
4768 | } | |
4769 | return; | |
4770 | ||
4771 | nla_put_failure: | |
4772 | nlmsg_free(skb); | |
4773 | failed: | |
4774 | drbd_err(peer_device, "Error %d while broadcasting event. Event seq:%u\n", | |
4775 | err, seq); | |
4776 | } | |
4777 | ||
4778 | void notify_helper(enum drbd_notification_type type, | |
4779 | struct drbd_device *device, struct drbd_connection *connection, | |
4780 | const char *name, int status) | |
4781 | { | |
4782 | struct drbd_resource *resource = device ? device->resource : connection->resource; | |
4783 | struct drbd_helper_info helper_info; | |
4784 | unsigned int seq = atomic_inc_return(¬ify_genl_seq); | |
4785 | struct sk_buff *skb = NULL; | |
4786 | struct drbd_genlmsghdr *dh; | |
4787 | int err; | |
4788 | ||
4789 | strlcpy(helper_info.helper_name, name, sizeof(helper_info.helper_name)); | |
4790 | helper_info.helper_name_len = min(strlen(name), sizeof(helper_info.helper_name)); | |
4791 | helper_info.helper_status = status; | |
4792 | ||
4793 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4794 | err = -ENOMEM; | |
4795 | if (!skb) | |
4796 | goto fail; | |
4797 | ||
4798 | err = -EMSGSIZE; | |
4799 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_HELPER); | |
4800 | if (!dh) | |
4801 | goto fail; | |
4802 | dh->minor = device ? device->minor : -1; | |
4803 | dh->ret_code = NO_ERROR; | |
4804 | mutex_lock(¬ification_mutex); | |
4805 | if (nla_put_drbd_cfg_context(skb, resource, connection, device) || | |
4806 | nla_put_notification_header(skb, type) || | |
4807 | drbd_helper_info_to_skb(skb, &helper_info, true)) | |
4808 | goto unlock_fail; | |
4809 | genlmsg_end(skb, dh); | |
d38f8612 | 4810 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4811 | skb = NULL; |
4812 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4813 | if (err && err != -ESRCH) | |
4814 | goto unlock_fail; | |
4815 | mutex_unlock(¬ification_mutex); | |
4816 | return; | |
4817 | ||
4818 | unlock_fail: | |
4819 | mutex_unlock(¬ification_mutex); | |
4820 | fail: | |
4821 | nlmsg_free(skb); | |
4822 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4823 | err, seq); | |
4824 | } | |
4825 | ||
4826 | static void notify_initial_state_done(struct sk_buff *skb, unsigned int seq) | |
4827 | { | |
4828 | struct drbd_genlmsghdr *dh; | |
4829 | int err; | |
4830 | ||
4831 | err = -EMSGSIZE; | |
4832 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_INITIAL_STATE_DONE); | |
4833 | if (!dh) | |
4834 | goto nla_put_failure; | |
4835 | dh->minor = -1U; | |
4836 | dh->ret_code = NO_ERROR; | |
4837 | if (nla_put_notification_header(skb, NOTIFY_EXISTS)) | |
4838 | goto nla_put_failure; | |
4839 | genlmsg_end(skb, dh); | |
4840 | return; | |
4841 | ||
4842 | nla_put_failure: | |
4843 | nlmsg_free(skb); | |
4844 | pr_err("Error %d sending event. Event seq:%u\n", err, seq); | |
4845 | } | |
4846 | ||
4847 | static void free_state_changes(struct list_head *list) | |
4848 | { | |
4849 | while (!list_empty(list)) { | |
4850 | struct drbd_state_change *state_change = | |
4851 | list_first_entry(list, struct drbd_state_change, list); | |
4852 | list_del(&state_change->list); | |
4853 | forget_state_change(state_change); | |
4854 | } | |
4855 | } | |
4856 | ||
4857 | static unsigned int notifications_for_state_change(struct drbd_state_change *state_change) | |
4858 | { | |
4859 | return 1 + | |
4860 | state_change->n_connections + | |
4861 | state_change->n_devices + | |
4862 | state_change->n_devices * state_change->n_connections; | |
4863 | } | |
4864 | ||
4865 | static int get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4866 | { | |
4867 | struct drbd_state_change *state_change = (struct drbd_state_change *)cb->args[0]; | |
4868 | unsigned int seq = cb->args[2]; | |
4869 | unsigned int n; | |
4870 | enum drbd_notification_type flags = 0; | |
4871 | ||
4872 | /* There is no need for taking notification_mutex here: it doesn't | |
4873 | matter if the initial state events mix with later state chage | |
4874 | events; we can always tell the events apart by the NOTIFY_EXISTS | |
4875 | flag. */ | |
4876 | ||
4877 | cb->args[5]--; | |
4878 | if (cb->args[5] == 1) { | |
4879 | notify_initial_state_done(skb, seq); | |
4880 | goto out; | |
4881 | } | |
4882 | n = cb->args[4]++; | |
4883 | if (cb->args[4] < cb->args[3]) | |
4884 | flags |= NOTIFY_CONTINUES; | |
4885 | if (n < 1) { | |
4886 | notify_resource_state_change(skb, seq, state_change->resource, | |
4887 | NOTIFY_EXISTS | flags); | |
4888 | goto next; | |
4889 | } | |
4890 | n--; | |
4891 | if (n < state_change->n_connections) { | |
4892 | notify_connection_state_change(skb, seq, &state_change->connections[n], | |
4893 | NOTIFY_EXISTS | flags); | |
4894 | goto next; | |
4895 | } | |
4896 | n -= state_change->n_connections; | |
4897 | if (n < state_change->n_devices) { | |
4898 | notify_device_state_change(skb, seq, &state_change->devices[n], | |
4899 | NOTIFY_EXISTS | flags); | |
4900 | goto next; | |
4901 | } | |
4902 | n -= state_change->n_devices; | |
4903 | if (n < state_change->n_devices * state_change->n_connections) { | |
4904 | notify_peer_device_state_change(skb, seq, &state_change->peer_devices[n], | |
4905 | NOTIFY_EXISTS | flags); | |
4906 | goto next; | |
4907 | } | |
4908 | ||
4909 | next: | |
4910 | if (cb->args[4] == cb->args[3]) { | |
4911 | struct drbd_state_change *next_state_change = | |
4912 | list_entry(state_change->list.next, | |
4913 | struct drbd_state_change, list); | |
4914 | cb->args[0] = (long)next_state_change; | |
4915 | cb->args[3] = notifications_for_state_change(next_state_change); | |
4916 | cb->args[4] = 0; | |
4917 | } | |
4918 | out: | |
4919 | return skb->len; | |
4920 | } | |
4921 | ||
4922 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4923 | { | |
4924 | struct drbd_resource *resource; | |
4925 | LIST_HEAD(head); | |
4926 | ||
4927 | if (cb->args[5] >= 1) { | |
4928 | if (cb->args[5] > 1) | |
4929 | return get_initial_state(skb, cb); | |
4930 | if (cb->args[0]) { | |
4931 | struct drbd_state_change *state_change = | |
4932 | (struct drbd_state_change *)cb->args[0]; | |
4933 | ||
4934 | /* connect list to head */ | |
4935 | list_add(&head, &state_change->list); | |
4936 | free_state_changes(&head); | |
4937 | } | |
4938 | return 0; | |
4939 | } | |
4940 | ||
4941 | cb->args[5] = 2; /* number of iterations */ | |
4942 | mutex_lock(&resources_mutex); | |
4943 | for_each_resource(resource, &drbd_resources) { | |
4944 | struct drbd_state_change *state_change; | |
4945 | ||
4946 | state_change = remember_old_state(resource, GFP_KERNEL); | |
4947 | if (!state_change) { | |
4948 | if (!list_empty(&head)) | |
4949 | free_state_changes(&head); | |
4950 | mutex_unlock(&resources_mutex); | |
4951 | return -ENOMEM; | |
4952 | } | |
4953 | copy_old_to_new_state_change(state_change); | |
4954 | list_add_tail(&state_change->list, &head); | |
4955 | cb->args[5] += notifications_for_state_change(state_change); | |
4956 | } | |
4957 | mutex_unlock(&resources_mutex); | |
4958 | ||
4959 | if (!list_empty(&head)) { | |
4960 | struct drbd_state_change *state_change = | |
4961 | list_entry(head.next, struct drbd_state_change, list); | |
4962 | cb->args[0] = (long)state_change; | |
4963 | cb->args[3] = notifications_for_state_change(state_change); | |
4964 | list_del(&head); /* detach list from head */ | |
4965 | } | |
4966 | ||
4967 | cb->args[2] = cb->nlh->nlmsg_seq; | |
4968 | return get_initial_state(skb, cb); | |
4969 | } |